diff --git "a/adcom-flux-klein-4b-lora-training.ipynb" "b/adcom-flux-klein-4b-lora-training.ipynb"
--- "a/adcom-flux-klein-4b-lora-training.ipynb"
+++ "b/adcom-flux-klein-4b-lora-training.ipynb"
@@ -1 +1 @@
-{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.12.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":14564379,"sourceType":"datasetVersion","datasetId":8022630}],"dockerImageVersionId":31260,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"HOW TO USE:\n1) Upload a civitai dataset .zip file to your google drive named kaggleset.zip \n2) Use the https://huggingface.co/datasets/codeShare/lora-training-data/blob/main/civit_dataset_to_latent.ipynb notebook \nto convert this dataset to flux_captions.json and flux_latents.safetensors (saved to your drive upon running the script)\n3) Create a private dataset called image-caption-dataset\n4) Add the flux_captions.json and flux_latents.safetensor to this dataset\n5) In this notebook , press the '+ Add input' button and select your private dataset\n6) Run this notebook\n//----//\nIf you have ideas on improvements / developments on FLUX Klein 4B LoRa \ntraining let me know in the comment section of this repo","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"#CELL 1\n!pip uninstall -y torch torchvision torchaudio diffusers accelerate peft transformers\n\n!pip install --no-deps torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu121\n\n!pip install --upgrade --no-cache-dir diffusers transformers accelerate peft safetensors tqdm huggingface-hub\n\n!pip install git+https://github.com/huggingface/diffusers.git","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:41:12.253502Z","iopub.execute_input":"2026-01-21T06:41:12.253751Z","iopub.status.idle":"2026-01-21T06:44:24.039967Z","shell.execute_reply.started":"2026-01-21T06:41:12.253730Z","shell.execute_reply":"2026-01-21T06:44:24.039276Z"}},"outputs":[{"name":"stdout","text":"Found existing installation: torch 2.8.0+cu126\nUninstalling torch-2.8.0+cu126:\n  Successfully uninstalled torch-2.8.0+cu126\nFound existing installation: torchvision 0.23.0+cu126\nUninstalling torchvision-0.23.0+cu126:\n  Successfully uninstalled torchvision-0.23.0+cu126\nFound existing installation: torchaudio 2.8.0+cu126\nUninstalling torchaudio-2.8.0+cu126:\n  Successfully uninstalled torchaudio-2.8.0+cu126\nFound existing installation: diffusers 0.35.2\nUninstalling diffusers-0.35.2:\n  Successfully uninstalled diffusers-0.35.2\nFound existing installation: accelerate 1.11.0\nUninstalling accelerate-1.11.0:\n  Successfully uninstalled accelerate-1.11.0\nFound existing installation: peft 0.17.1\nUninstalling peft-0.17.1:\n  Successfully uninstalled peft-0.17.1\nFound existing installation: transformers 4.57.1\nUninstalling transformers-4.57.1:\n  Successfully uninstalled transformers-4.57.1\nLooking in indexes: https://download.pytorch.org/whl/cu121\nCollecting torch==2.5.1\n  Downloading https://download.pytorch.org/whl/cu121/torch-2.5.1%2Bcu121-cp312-cp312-linux_x86_64.whl (780.4 MB)\n\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m780.4/780.4 MB\u001b[0m \u001b[31m2.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hCollecting torchvision==0.20.1\n  Downloading https://download.pytorch.org/whl/cu121/torchvision-0.20.1%2Bcu121-cp312-cp312-linux_x86_64.whl (7.3 MB)\n\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.3/7.3 MB\u001b[0m \u001b[31m122.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m\n\u001b[?25hCollecting torchaudio==2.5.1\n  Downloading https://download.pytorch.org/whl/cu121/torchaudio-2.5.1%2Bcu121-cp312-cp312-linux_x86_64.whl (3.4 MB)\n\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m3.4/3.4 MB\u001b[0m \u001b[31m95.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m\n\u001b[?25hInstalling collected packages: torchaudio, torchvision, torch\nSuccessfully installed torch-2.5.1+cu121 torchaudio-2.5.1+cu121 torchvision-0.20.1+cu121\nCollecting diffusers\n  Downloading diffusers-0.36.0-py3-none-any.whl.metadata (20 kB)\nCollecting transformers\n  Downloading transformers-4.57.6-py3-none-any.whl.metadata (43 kB)\n\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m44.0/44.0 kB\u001b[0m \u001b[31m7.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hCollecting accelerate\n  Downloading accelerate-1.12.0-py3-none-any.whl.metadata (19 kB)\nCollecting peft\n  Downloading peft-0.18.1-py3-none-any.whl.metadata (14 kB)\nRequirement already satisfied: safetensors in /usr/local/lib/python3.12/dist-packages (0.6.2)\nCollecting safetensors\n  Downloading safetensors-0.7.0-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (4.1 kB)\nRequirement already satisfied: tqdm in /usr/local/lib/python3.12/dist-packages (4.67.1)\nRequirement already satisfied: huggingface-hub in /usr/local/lib/python3.12/dist-packages (0.36.0)\nCollecting huggingface-hub\n  Downloading huggingface_hub-1.3.2-py3-none-any.whl.metadata (13 kB)\nRequirement already satisfied: importlib_metadata in /usr/local/lib/python3.12/dist-packages (from diffusers) (8.7.0)\nRequirement already satisfied: filelock in /usr/local/lib/python3.12/dist-packages (from diffusers) (3.20.3)\nRequirement already satisfied: httpx<1.0.0 in /usr/local/lib/python3.12/dist-packages (from diffusers) (0.28.1)\nRequirement already satisfied: numpy in /usr/local/lib/python3.12/dist-packages (from diffusers) (2.0.2)\nRequirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.12/dist-packages (from diffusers) (2025.11.3)\nRequirement already satisfied: requests in /usr/local/lib/python3.12/dist-packages (from diffusers) (2.32.5)\nRequirement already satisfied: Pillow in /usr/local/lib/python3.12/dist-packages (from diffusers) (11.3.0)\nRequirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.12/dist-packages (from transformers) (26.0rc2)\nRequirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.12/dist-packages (from transformers) (6.0.3)\nRequirement already satisfied: tokenizers<=0.23.0,>=0.22.0 in /usr/local/lib/python3.12/dist-packages (from transformers) (0.22.1)\nRequirement already satisfied: psutil in /usr/local/lib/python3.12/dist-packages (from accelerate) (5.9.5)\nRequirement already satisfied: torch>=2.0.0 in /usr/local/lib/python3.12/dist-packages (from accelerate) (2.5.1+cu121)\nRequirement already satisfied: fsspec>=2023.5.0 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub) (2025.10.0)\nRequirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub) (4.15.0)\nRequirement already satisfied: hf-xet<2.0.0,>=1.1.3 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub) (1.2.1rc0)\nRequirement already satisfied: anyio in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers) (4.12.1)\nRequirement already satisfied: certifi in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers) (2026.1.4)\nRequirement already satisfied: httpcore==1.* in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers) (1.0.9)\nRequirement already satisfied: idna in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers) (3.11)\nRequirement already satisfied: h11>=0.16 in /usr/local/lib/python3.12/dist-packages (from httpcore==1.*->httpx<1.0.0->diffusers) (0.16.0)\nRequirement already satisfied: networkx in /usr/local/lib/python3.12/dist-packages (from torch>=2.0.0->accelerate) (3.5)\nRequirement already satisfied: jinja2 in /usr/local/lib/python3.12/dist-packages (from torch>=2.0.0->accelerate) (3.1.6)\nCollecting nvidia-cuda-nvrtc-cu12==12.1.105 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cuda_nvrtc_cu12-12.1.105-py3-none-manylinux1_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-cuda-runtime-cu12==12.1.105 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cuda_runtime_cu12-12.1.105-py3-none-manylinux1_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-cuda-cupti-cu12==12.1.105 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cuda_cupti_cu12-12.1.105-py3-none-manylinux1_x86_64.whl.metadata (1.6 kB)\nCollecting nvidia-cudnn-cu12==9.1.0.70 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cudnn_cu12-9.1.0.70-py3-none-manylinux2014_x86_64.whl.metadata (1.6 kB)\nCollecting nvidia-cublas-cu12==12.1.3.1 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cublas_cu12-12.1.3.1-py3-none-manylinux1_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-cufft-cu12==11.0.2.54 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cufft_cu12-11.0.2.54-py3-none-manylinux1_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-curand-cu12==10.3.2.106 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_curand_cu12-10.3.2.106-py3-none-manylinux1_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-cusolver-cu12==11.4.5.107 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cusolver_cu12-11.4.5.107-py3-none-manylinux1_x86_64.whl.metadata (1.6 kB)\nCollecting nvidia-cusparse-cu12==12.1.0.106 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_cusparse_cu12-12.1.0.106-py3-none-manylinux1_x86_64.whl.metadata (1.6 kB)\nCollecting nvidia-nccl-cu12==2.21.5 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_nccl_cu12-2.21.5-py3-none-manylinux2014_x86_64.whl.metadata (1.8 kB)\nCollecting nvidia-nvtx-cu12==12.1.105 (from torch>=2.0.0->accelerate)\n  Downloading nvidia_nvtx_cu12-12.1.105-py3-none-manylinux1_x86_64.whl.metadata (1.7 kB)\nCollecting triton==3.1.0 (from torch>=2.0.0->accelerate)\n  Downloading triton-3.1.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (1.3 kB)\nRequirement already satisfied: setuptools in /usr/local/lib/python3.12/dist-packages (from torch>=2.0.0->accelerate) (75.2.0)\nCollecting sympy==1.13.1 (from torch>=2.0.0->accelerate)\n  Downloading sympy-1.13.1-py3-none-any.whl.metadata (12 kB)\nRequirement already satisfied: nvidia-nvjitlink-cu12 in /usr/local/lib/python3.12/dist-packages (from nvidia-cusolver-cu12==11.4.5.107->torch>=2.0.0->accelerate) (12.6.85)\nRequirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.12/dist-packages (from sympy==1.13.1->torch>=2.0.0->accelerate) (1.3.0)\nRequirement already satisfied: zipp>=3.20 in /usr/local/lib/python3.12/dist-packages (from importlib_metadata->diffusers) (3.23.0)\nRequirement already satisfied: charset_normalizer<4,>=2 in /usr/local/lib/python3.12/dist-packages (from requests->diffusers) (3.4.4)\nRequirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.12/dist-packages (from requests->diffusers) (2.6.3)\nRequirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.12/dist-packages (from jinja2->torch>=2.0.0->accelerate) (3.0.3)\nDownloading diffusers-0.36.0-py3-none-any.whl (4.6 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m4.6/4.6 MB\u001b[0m \u001b[31m70.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading transformers-4.57.6-py3-none-any.whl (12.0 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m12.0/12.0 MB\u001b[0m \u001b[31m209.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading accelerate-1.12.0-py3-none-any.whl (380 kB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m380.9/380.9 kB\u001b[0m \u001b[31m154.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading peft-0.18.1-py3-none-any.whl (556 kB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m557.0/557.0 kB\u001b[0m \u001b[31m376.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading safetensors-0.7.0-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (507 kB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m507.2/507.2 kB\u001b[0m \u001b[31m199.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading nvidia_cublas_cu12-12.1.3.1-py3-none-manylinux1_x86_64.whl (410.6 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m410.6/410.6 MB\u001b[0m \u001b[31m265.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cuda_cupti_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (14.1 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m14.1/14.1 MB\u001b[0m \u001b[31m252.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cuda_nvrtc_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (23.7 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m23.7/23.7 MB\u001b[0m \u001b[31m177.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cuda_runtime_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (823 kB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m823.6/823.6 kB\u001b[0m \u001b[31m303.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading nvidia_cudnn_cu12-9.1.0.70-py3-none-manylinux2014_x86_64.whl (664.8 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m664.8/664.8 MB\u001b[0m \u001b[31m251.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cufft_cu12-11.0.2.54-py3-none-manylinux1_x86_64.whl (121.6 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m121.6/121.6 MB\u001b[0m \u001b[31m282.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading nvidia_curand_cu12-10.3.2.106-py3-none-manylinux1_x86_64.whl (56.5 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m56.5/56.5 MB\u001b[0m \u001b[31m211.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cusolver_cu12-11.4.5.107-py3-none-manylinux1_x86_64.whl (124.2 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m124.2/124.2 MB\u001b[0m \u001b[31m275.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cusparse_cu12-12.1.0.106-py3-none-manylinux1_x86_64.whl (196.0 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m196.0/196.0 MB\u001b[0m \u001b[31m198.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading nvidia_nccl_cu12-2.21.5-py3-none-manylinux2014_x86_64.whl (188.7 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m188.7/188.7 MB\u001b[0m \u001b[31m147.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_nvtx_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (99 kB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m99.1/99.1 kB\u001b[0m \u001b[31m320.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading sympy-1.13.1-py3-none-any.whl (6.2 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m6.2/6.2 MB\u001b[0m \u001b[31m153.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n\u001b[?25hDownloading triton-3.1.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (209.6 MB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m209.6/209.6 MB\u001b[0m \u001b[31m158.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hInstalling collected packages: triton, sympy, safetensors, nvidia-nvtx-cu12, nvidia-nccl-cu12, nvidia-cusparse-cu12, nvidia-curand-cu12, nvidia-cufft-cu12, nvidia-cuda-runtime-cu12, nvidia-cuda-nvrtc-cu12, nvidia-cuda-cupti-cu12, nvidia-cublas-cu12, nvidia-cusolver-cu12, nvidia-cudnn-cu12, diffusers, transformers, accelerate, peft\n  Attempting uninstall: triton\n    Found existing installation: triton 3.4.0\n    Uninstalling triton-3.4.0:\n      Successfully uninstalled triton-3.4.0\n  Attempting uninstall: sympy\n    Found existing installation: sympy 1.13.3\n    Uninstalling sympy-1.13.3:\n      Successfully uninstalled sympy-1.13.3\n  Attempting uninstall: safetensors\n    Found existing installation: safetensors 0.6.2\n    Uninstalling safetensors-0.6.2:\n      Successfully uninstalled safetensors-0.6.2\n  Attempting uninstall: nvidia-nvtx-cu12\n    Found existing installation: nvidia-nvtx-cu12 12.6.77\n    Uninstalling nvidia-nvtx-cu12-12.6.77:\n      Successfully uninstalled nvidia-nvtx-cu12-12.6.77\n  Attempting uninstall: nvidia-nccl-cu12\n    Found existing installation: nvidia-nccl-cu12 2.27.3\n    Uninstalling nvidia-nccl-cu12-2.27.3:\n      Successfully uninstalled nvidia-nccl-cu12-2.27.3\n  Attempting uninstall: nvidia-cusparse-cu12\n    Found existing installation: nvidia-cusparse-cu12 12.5.4.2\n    Uninstalling nvidia-cusparse-cu12-12.5.4.2:\n      Successfully uninstalled nvidia-cusparse-cu12-12.5.4.2\n  Attempting uninstall: nvidia-curand-cu12\n    Found existing installation: nvidia-curand-cu12 10.3.7.77\n    Uninstalling nvidia-curand-cu12-10.3.7.77:\n      Successfully uninstalled nvidia-curand-cu12-10.3.7.77\n  Attempting uninstall: nvidia-cufft-cu12\n    Found existing installation: nvidia-cufft-cu12 11.3.0.4\n    Uninstalling nvidia-cufft-cu12-11.3.0.4:\n      Successfully uninstalled nvidia-cufft-cu12-11.3.0.4\n  Attempting uninstall: nvidia-cuda-runtime-cu12\n    Found existing installation: nvidia-cuda-runtime-cu12 12.6.77\n    Uninstalling nvidia-cuda-runtime-cu12-12.6.77:\n      Successfully uninstalled nvidia-cuda-runtime-cu12-12.6.77\n  Attempting uninstall: nvidia-cuda-nvrtc-cu12\n    Found existing installation: nvidia-cuda-nvrtc-cu12 12.6.77\n    Uninstalling nvidia-cuda-nvrtc-cu12-12.6.77:\n      Successfully uninstalled nvidia-cuda-nvrtc-cu12-12.6.77\n  Attempting uninstall: nvidia-cuda-cupti-cu12\n    Found existing installation: nvidia-cuda-cupti-cu12 12.6.80\n    Uninstalling nvidia-cuda-cupti-cu12-12.6.80:\n      Successfully uninstalled nvidia-cuda-cupti-cu12-12.6.80\n  Attempting uninstall: nvidia-cublas-cu12\n    Found existing installation: nvidia-cublas-cu12 12.6.4.1\n    Uninstalling nvidia-cublas-cu12-12.6.4.1:\n      Successfully uninstalled nvidia-cublas-cu12-12.6.4.1\n  Attempting uninstall: nvidia-cusolver-cu12\n    Found existing installation: nvidia-cusolver-cu12 11.7.1.2\n    Uninstalling nvidia-cusolver-cu12-11.7.1.2:\n      Successfully uninstalled nvidia-cusolver-cu12-11.7.1.2\n  Attempting uninstall: nvidia-cudnn-cu12\n    Found existing installation: nvidia-cudnn-cu12 9.10.2.21\n    Uninstalling nvidia-cudnn-cu12-9.10.2.21:\n      Successfully uninstalled nvidia-cudnn-cu12-9.10.2.21\n\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\ncudf-cu12 25.6.0 requires pyarrow<20.0.0a0,>=14.0.0; platform_machine == \"x86_64\", but you have pyarrow 22.0.0 which is incompatible.\nfastai 2.8.4 requires fastcore<1.9,>=1.8.0, but you have fastcore 1.11.3 which is incompatible.\u001b[0m\u001b[31m\n\u001b[0mSuccessfully installed accelerate-1.12.0 diffusers-0.36.0 nvidia-cublas-cu12-12.1.3.1 nvidia-cuda-cupti-cu12-12.1.105 nvidia-cuda-nvrtc-cu12-12.1.105 nvidia-cuda-runtime-cu12-12.1.105 nvidia-cudnn-cu12-9.1.0.70 nvidia-cufft-cu12-11.0.2.54 nvidia-curand-cu12-10.3.2.106 nvidia-cusolver-cu12-11.4.5.107 nvidia-cusparse-cu12-12.1.0.106 nvidia-nccl-cu12-2.21.5 nvidia-nvtx-cu12-12.1.105 peft-0.18.1 safetensors-0.7.0 sympy-1.13.1 transformers-4.57.6 triton-3.1.0\nCollecting git+https://github.com/huggingface/diffusers.git\n  Cloning https://github.com/huggingface/diffusers.git to /tmp/pip-req-build-5qs94bpu\n  Running command git clone --filter=blob:none --quiet https://github.com/huggingface/diffusers.git /tmp/pip-req-build-5qs94bpu\n  Resolved https://github.com/huggingface/diffusers.git to commit ec376293714f269947f6d9d8a572bd73040bc1a0\n  Installing build dependencies ... \u001b[?25l\u001b[?25hdone\n  Getting requirements to build wheel ... \u001b[?25l\u001b[?25hdone\n  Preparing metadata (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\nRequirement already satisfied: importlib_metadata in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (8.7.0)\nRequirement already satisfied: filelock in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (3.20.3)\nRequirement already satisfied: httpx<1.0.0 in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (0.28.1)\nRequirement already satisfied: huggingface-hub<2.0,>=0.34.0 in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (0.36.0)\nRequirement already satisfied: numpy in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (2.0.2)\nRequirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (2025.11.3)\nRequirement already satisfied: requests in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (2.32.5)\nRequirement already satisfied: safetensors>=0.3.1 in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (0.7.0)\nRequirement already satisfied: Pillow in /usr/local/lib/python3.12/dist-packages (from diffusers==0.37.0.dev0) (11.3.0)\nRequirement already satisfied: anyio in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers==0.37.0.dev0) (4.12.1)\nRequirement already satisfied: certifi in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers==0.37.0.dev0) (2026.1.4)\nRequirement already satisfied: httpcore==1.* in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers==0.37.0.dev0) (1.0.9)\nRequirement already satisfied: idna in /usr/local/lib/python3.12/dist-packages (from httpx<1.0.0->diffusers==0.37.0.dev0) (3.11)\nRequirement already satisfied: h11>=0.16 in /usr/local/lib/python3.12/dist-packages (from httpcore==1.*->httpx<1.0.0->diffusers==0.37.0.dev0) (0.16.0)\nRequirement already satisfied: fsspec>=2023.5.0 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=0.34.0->diffusers==0.37.0.dev0) (2025.10.0)\nRequirement already satisfied: packaging>=20.9 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=0.34.0->diffusers==0.37.0.dev0) (26.0rc2)\nRequirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=0.34.0->diffusers==0.37.0.dev0) (6.0.3)\nRequirement already satisfied: tqdm>=4.42.1 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=0.34.0->diffusers==0.37.0.dev0) (4.67.1)\nRequirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=0.34.0->diffusers==0.37.0.dev0) (4.15.0)\nRequirement already satisfied: hf-xet<2.0.0,>=1.1.3 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=0.34.0->diffusers==0.37.0.dev0) (1.2.1rc0)\nRequirement already satisfied: zipp>=3.20 in /usr/local/lib/python3.12/dist-packages (from importlib_metadata->diffusers==0.37.0.dev0) (3.23.0)\nRequirement already satisfied: charset_normalizer<4,>=2 in /usr/local/lib/python3.12/dist-packages (from requests->diffusers==0.37.0.dev0) (3.4.4)\nRequirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.12/dist-packages (from requests->diffusers==0.37.0.dev0) (2.6.3)\nBuilding wheels for collected packages: diffusers\n  Building wheel for diffusers (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\n  Created wheel for diffusers: filename=diffusers-0.37.0.dev0-py3-none-any.whl size=4893406 sha256=76089d2f822b7c1086ff1fd07ba58a03f82c6b49ec2ea569ea3596248d511089\n  Stored in directory: /tmp/pip-ephem-wheel-cache-3f4dmw0k/wheels/23/0f/7d/f97813d265ed0e599a78d83afd4e1925740896ca79b46cccfd\nSuccessfully built diffusers\nInstalling collected packages: diffusers\n  Attempting uninstall: diffusers\n    Found existing installation: diffusers 0.36.0\n    Uninstalling diffusers-0.36.0:\n      Successfully uninstalled diffusers-0.36.0\nSuccessfully installed diffusers-0.37.0.dev0\n","output_type":"stream"}],"execution_count":1},{"cell_type":"code","source":"# CELL 2 — Verify\nimport torch, diffusers\n\nprint(\"Torch:\", torch.__version__)\nprint(\"Diffusers:\", diffusers.__version__)\nprint(\"CUDA:\", torch.cuda.is_available())\nprint(\"GPU:\", torch.cuda.get_device_name(0) if torch.cuda.is_available() else \"None\")\n\nfrom diffusers import Flux2KleinPipeline\nprint(\"Flux2KleinPipeline OK\")\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:45:52.733371Z","iopub.execute_input":"2026-01-21T06:45:52.733920Z","iopub.status.idle":"2026-01-21T06:45:52.738746Z","shell.execute_reply.started":"2026-01-21T06:45:52.733890Z","shell.execute_reply":"2026-01-21T06:45:52.738076Z"}},"outputs":[{"name":"stdout","text":"Torch: 2.5.1+cu121\nDiffusers: 0.37.0.dev0\nCUDA: True\nGPU: Tesla P100-PCIE-16GB\nFlux2KleinPipeline OK\n","output_type":"stream"}],"execution_count":4},{"cell_type":"code","source":"# CELL 3 — Config\nimport os\n\ndevice = \"cuda\"\ndtype  = torch.float16\n\nDATASET_NAME = \"image-caption-dataset\"\n\nCAPTIONS_PATH = f\"/kaggle/input/{DATASET_NAME}/flux_captions.json\"\nLATENTS_PATH  = f\"/kaggle/input/{DATASET_NAME}/flux_latents.safetensors\"\n\nCACHE_DIR = \"/kaggle/working/cache\"\nSAVE_DIR  = \"/kaggle/working/flux_klein_lora\"\n\nos.makedirs(CACHE_DIR, exist_ok=True)\nos.makedirs(SAVE_DIR, exist_ok=True)\n\n# training\nRANK = 16\nALPHA = 16\nLR = 2e-5\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:45:55.735789Z","iopub.execute_input":"2026-01-21T06:45:55.736570Z","iopub.status.idle":"2026-01-21T06:45:55.741313Z","shell.execute_reply.started":"2026-01-21T06:45:55.736539Z","shell.execute_reply":"2026-01-21T06:45:55.740544Z"}},"outputs":[],"execution_count":5},{"cell_type":"code","source":"# CELL 4 — Load captions + latents\nimport json\nfrom safetensors.torch import load_file\n\nwith open(CAPTIONS_PATH) as f:\n    captions = json.load(f)\n\nlatents = load_file(LATENTS_PATH)\n\nkeys = list(captions.keys())\nprint(\"Samples:\", len(keys))\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:45:58.248254Z","iopub.execute_input":"2026-01-21T06:45:58.248884Z","iopub.status.idle":"2026-01-21T06:45:58.701994Z","shell.execute_reply.started":"2026-01-21T06:45:58.248854Z","shell.execute_reply":"2026-01-21T06:45:58.701386Z"}},"outputs":[{"name":"stdout","text":"Samples: 125\n","output_type":"stream"}],"execution_count":6},{"cell_type":"code","source":"# CELL 5 — Dataset (returns latent + key)\nimport torch\nfrom torch.utils.data import Dataset, DataLoader\n\nclass FluxLatentDataset(Dataset):\n    def __len__(self):\n        return len(keys)\n\n    def __getitem__(self, idx):\n        k = keys[idx]\n        return latents[k], k\n\ndataset = FluxLatentDataset()\nloader = DataLoader(dataset, batch_size=1, shuffle=True)\n\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:46:01.404819Z","iopub.execute_input":"2026-01-21T06:46:01.405424Z","iopub.status.idle":"2026-01-21T06:46:01.410301Z","shell.execute_reply.started":"2026-01-21T06:46:01.405394Z","shell.execute_reply":"2026-01-21T06:46:01.409594Z"}},"outputs":[],"execution_count":7},{"cell_type":"code","source":"# CELL 6 — Encode text on GPU and CACHE FLUX-READY embeddings\nimport torch, gc\nfrom transformers import AutoTokenizer, AutoModel\n\nMODEL_ID = \"black-forest-labs/FLUX.2-klein-4B\"\n\ntokenizer = AutoTokenizer.from_pretrained(\n    MODEL_ID,\n    subfolder=\"tokenizer\",\n    trust_remote_code=True,\n    cache_dir=CACHE_DIR,\n)\n\ntext_encoder = AutoModel.from_pretrained(\n    MODEL_ID,\n    subfolder=\"text_encoder\",\n    trust_remote_code=True,\n    dtype=torch.float16,\n    cache_dir=CACHE_DIR,\n).to(\"cuda\")\n\ntext_encoder.eval()\n\ntext_cache = {}\n\nwith torch.no_grad():\n    for k, caption in captions.items():\n        inputs = tokenizer(\n            caption,\n            padding=\"max_length\",\n            truncation=True,\n            max_length=128,\n            return_tensors=\"pt\"\n        ).to(\"cuda\")\n\n        outputs = text_encoder(**inputs, output_hidden_states=True, return_dict=True)\n        txt = outputs.hidden_states[-1]          # [1, T, 2560]\n        txt = txt.repeat(1, 1, 3)                # → [1, T, 7680]\n        text_cache[k] = txt.cpu()\n\nprint(\"✅ Cached FLUX-ready text embeddings.\")\n\ndel text_encoder\ntorch.cuda.empty_cache()\ngc.collect()\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:46:04.538956Z","iopub.execute_input":"2026-01-21T06:46:04.539598Z","iopub.status.idle":"2026-01-21T06:47:02.088127Z","shell.execute_reply.started":"2026-01-21T06:46:04.539569Z","shell.execute_reply":"2026-01-21T06:47:02.087363Z"}},"outputs":[{"output_type":"display_data","data":{"text/plain":"tokenizer_config.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"7eb47566c5b642a0a9e8049c67bfd562"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"vocab.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"8a7ffb1cb3f34f1cb20db3369301886c"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"merges.txt: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"5648fc23b5e448f0a002da76c7a2cc67"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"tokenizer/tokenizer.json:   0%|          | 0.00/11.4M [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"0504e1ee64724b409c527c0266cf5864"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"added_tokens.json:   0%|          | 0.00/707 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"1cd3f580cebb43689d537c76b99b5517"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"special_tokens_map.json:   0%|          | 0.00/613 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"68fa4dacfd414607acc254a7001fe9cb"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"chat_template.jinja: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"a619e3c684394997ac25522ee88aed23"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"config.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"02e0198de1164fd4a5fd0be433b136e4"}},"metadata":{}},{"name":"stderr","text":"`torch_dtype` is deprecated! Use `dtype` instead!\n","output_type":"stream"},{"output_type":"display_data","data":{"text/plain":"model.safetensors.index.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"af52526aac28456ea10b96f98ddca3df"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Fetching 2 files:   0%|          | 0/2 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"f36c0d81d75c4c8f9c2956f645313817"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"text_encoder/model-00002-of-00002.safete(…):   0%|          | 0.00/3.08G [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"ee179610b92e4938883a48e47a455ed3"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"text_encoder/model-00001-of-00002.safete(…):   0%|          | 0.00/4.97G [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"0208a0c69faa4f9eb1747d8cd6fc2cea"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"6006b9a7aef34137b6d46074ec5c5ebf"}},"metadata":{}},{"name":"stdout","text":"✅ Cached FLUX-ready text embeddings.\n","output_type":"stream"},{"execution_count":8,"output_type":"execute_result","data":{"text/plain":"8834"},"metadata":{}}],"execution_count":8},{"cell_type":"code","source":"# =========================================================\n# CELL 7A — Inspect Klein Transformer Module Names\n# =========================================================\n\nnames = []\nfor name, module in pipe.transformer.named_modules():\n    if isinstance(module, torch.nn.Linear):\n        names.append(name)\n\n# print a sample\nfor n in names[:50]:\n    print(n)\n\nprint(\"\\n--- TOTAL Linear layers:\", len(names))\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:49:28.032293Z","iopub.execute_input":"2026-01-21T06:49:28.032618Z","iopub.status.idle":"2026-01-21T06:49:28.038556Z","shell.execute_reply.started":"2026-01-21T06:49:28.032589Z","shell.execute_reply":"2026-01-21T06:49:28.037744Z"}},"outputs":[{"name":"stdout","text":"time_guidance_embed.timestep_embedder.linear_1\ntime_guidance_embed.timestep_embedder.linear_2\ndouble_stream_modulation_img.linear\ndouble_stream_modulation_txt.linear\nsingle_stream_modulation.linear\nx_embedder\ncontext_embedder\ntransformer_blocks.0.attn.to_q\ntransformer_blocks.0.attn.to_k\ntransformer_blocks.0.attn.to_v\ntransformer_blocks.0.attn.to_out.0\ntransformer_blocks.0.attn.add_q_proj\ntransformer_blocks.0.attn.add_k_proj\ntransformer_blocks.0.attn.add_v_proj\ntransformer_blocks.0.attn.to_add_out\ntransformer_blocks.0.ff.linear_in\ntransformer_blocks.0.ff.linear_out\ntransformer_blocks.0.ff_context.linear_in\ntransformer_blocks.0.ff_context.linear_out\ntransformer_blocks.1.attn.to_q\ntransformer_blocks.1.attn.to_k\ntransformer_blocks.1.attn.to_v\ntransformer_blocks.1.attn.to_out.0\ntransformer_blocks.1.attn.add_q_proj\ntransformer_blocks.1.attn.add_k_proj\ntransformer_blocks.1.attn.add_v_proj\ntransformer_blocks.1.attn.to_add_out\ntransformer_blocks.1.ff.linear_in\ntransformer_blocks.1.ff.linear_out\ntransformer_blocks.1.ff_context.linear_in\ntransformer_blocks.1.ff_context.linear_out\ntransformer_blocks.2.attn.to_q\ntransformer_blocks.2.attn.to_k\ntransformer_blocks.2.attn.to_v\ntransformer_blocks.2.attn.to_out.0\ntransformer_blocks.2.attn.add_q_proj\ntransformer_blocks.2.attn.add_k_proj\ntransformer_blocks.2.attn.add_v_proj\ntransformer_blocks.2.attn.to_add_out\ntransformer_blocks.2.ff.linear_in\ntransformer_blocks.2.ff.linear_out\ntransformer_blocks.2.ff_context.linear_in\ntransformer_blocks.2.ff_context.linear_out\ntransformer_blocks.3.attn.to_q\ntransformer_blocks.3.attn.to_k\ntransformer_blocks.3.attn.to_v\ntransformer_blocks.3.attn.to_out.0\ntransformer_blocks.3.attn.add_q_proj\ntransformer_blocks.3.attn.add_k_proj\ntransformer_blocks.3.attn.add_v_proj\n\n--- TOTAL Linear layers: 109\n","output_type":"stream"}],"execution_count":10},{"cell_type":"code","source":"# =========================================================\n# CELL 7 — Load FLUX Transformer and Attach PROPER LoRA\n# =========================================================\n\nimport torch, gc\nfrom diffusers import Flux2KleinPipeline\nfrom peft import LoraConfig, get_peft_model\n\nMODEL_ID = \"black-forest-labs/FLUX.2-klein-4B\"\n\npipe = Flux2KleinPipeline.from_pretrained(\n    MODEL_ID,\n    torch_dtype=torch.float16,\n    cache_dir=CACHE_DIR,\n)\n\n# keep ONLY transformer\npipe.text_encoder = None\npipe.vae = None\npipe.scheduler = None\n\ntransformer = pipe.transformer\n\n# ---------------- LoRA TARGETS (FLUX-COMPATIBLE) ----------------\n\nTARGET_MODULES = [\n    \"attn.to_q\",\n    \"attn.to_k\",\n    \"attn.to_v\",\n    \"attn.to_out.0\",\n    \"attn.add_q_proj\",\n    \"attn.add_k_proj\",\n    \"attn.add_v_proj\",\n    \"attn.to_add_out\",\n    \"ff.linear_in\",\n    \"ff.linear_out\",\n    \"ff_context.linear_in\",\n    \"ff_context.linear_out\",\n]\n\nlora_config = LoraConfig(\n    r=RANK,\n    lora_alpha=ALPHA,\n    target_modules=TARGET_MODULES,\n    lora_dropout=0.05,\n    bias=\"none\",\n)\n\ntransformer = get_peft_model(transformer, lora_config)\ntransformer.enable_gradient_checkpointing()\n\npipe.transformer = transformer\n\ntrainable = sum(p.numel() for p in transformer.parameters() if p.requires_grad)\ntotal = sum(p.numel() for p in transformer.parameters())\n\nprint(f\"Trainable params: {trainable/1e6:.2f}M / {total/1e6:.2f}M\")\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:51:18.251124Z","iopub.execute_input":"2026-01-21T06:51:18.251459Z","iopub.status.idle":"2026-01-21T06:52:31.616038Z","shell.execute_reply.started":"2026-01-21T06:51:18.251433Z","shell.execute_reply":"2026-01-21T06:52:31.615093Z"}},"outputs":[{"output_type":"display_data","data":{"text/plain":"Loading pipeline components...:   0%|          | 0/5 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"81fced2a8a3c4ac3a83e0ec8ba021858"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"cd5a0cb701cb4f5f9c6bcbb8b202639b"}},"metadata":{}},{"name":"stdout","text":"Trainable params: 9.34M / 3884.88M\n","output_type":"stream"}],"execution_count":12},{"cell_type":"code","source":"# CELL 8 — Move Transformer to GPU Only\npipe.transformer = pipe.transformer.to(\"cuda\")\n\ngc.collect()\ntorch.cuda.empty_cache()\n\nprint(\"✅ Transformer on GPU.\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:52:31.618488Z","iopub.execute_input":"2026-01-21T06:52:31.618839Z","iopub.status.idle":"2026-01-21T06:52:34.734557Z","shell.execute_reply.started":"2026-01-21T06:52:31.618800Z","shell.execute_reply":"2026-01-21T06:52:34.733924Z"}},"outputs":[{"name":"stdout","text":"✅ Transformer on GPU.\n","output_type":"stream"}],"execution_count":13},{"cell_type":"code","source":"# =========================================================\n# CELL 9 — Flow Matching Training (Fixed Shapes, Safe Saves)\n# =========================================================\n\nimport torch, time, os\nimport torch.nn.functional as F\nfrom tqdm import tqdm\n\nfrom peft import get_peft_model_state_dict\nfrom safetensors.torch import save_file\n\n# ---------------- Config ----------------\n\nEPOCHS = 1\nSAVE_EVERY = 3\nFLOW_T_SCALE = 50.0\nCLAMP_VAL = 3.0\nGRAD_CLIP = 0.5\n\nMAX_SECONDS = (11 * 60 + 40) * 60   # 11h 40m\n\ndevice = \"cuda\"\ndtype = torch.float16\n\npipe.transformer.train()\n\n# ---------------- Helpers ----------------\n\ndef patchify_latents(latents):\n    assert latents.ndim == 4, f\"Expected [B,C,H,W], got {latents.shape}\"\n    B, C, H, W = latents.shape\n    assert H % 2 == 0 and W % 2 == 0\n\n    latents = latents.reshape(B, C, H//2, 2, W//2, 2)\n    latents = latents.permute(0, 2, 4, 1, 3, 5).contiguous()\n    return latents.view(B, (H//2)*(W//2), C*2*2)  # [B, 4096, 128]\n\n\ndef generate_flux_pos_ids(batch, ph, pw, txt_len, device, dtype):\n    y = torch.linspace(0, 1, ph, device=device, dtype=dtype)\n    x = torch.linspace(0, 1, pw, device=device, dtype=dtype)\n    gy, gx = torch.meshgrid(y, x, indexing=\"ij\")\n\n    pos_y = gy.flatten()\n    pos_x = gx.flatten()\n    zeros = torch.zeros_like(pos_y)\n\n    img_ids = torch.stack([pos_y, pos_x, zeros, zeros], dim=-1)\n    img_ids = img_ids.unsqueeze(0).repeat(batch, 1, 1)\n\n    t = torch.arange(txt_len, device=device, dtype=dtype) / txt_len\n    txt_ids = torch.stack(\n        [t, torch.zeros_like(t), torch.zeros_like(t), torch.zeros_like(t)], dim=-1\n    )\n    txt_ids = txt_ids.unsqueeze(0).repeat(batch, 1, 1)\n\n    return img_ids, txt_ids\n\n\n# ---------------- Optimizer ----------------\n\ntrainable_params = [p for p in pipe.transformer.parameters() if p.requires_grad]\n\noptimizer = torch.optim.AdamW(\n    trainable_params,\n    lr=LR,\n    betas=(0.9, 0.999),\n    weight_decay=1e-4,\n)\n\n# ---------------- Training ----------------\n\nstart_time = time.time()\n\nprint(\"⏱️ Training started. Max time:\", MAX_SECONDS / 3600, \"hours\")\n\nfor epoch in range(1, EPOCHS + 1):\n\n    print(f\"\\n===== Epoch {epoch} =====\")\n    epoch_loss = 0.0\n\n    for latent_b, key in tqdm(loader, desc=f\"Epoch {epoch}\"):\n\n        if time.time() - start_time >= MAX_SECONDS:\n            print(\"⏰ Time limit reached. Stopping training.\")\n            break\n\n        # ---- latent ----\n        latent_b = latent_b.to(device, dtype=dtype)\n\n        if latent_b.ndim == 5:\n            latent_b = latent_b.squeeze(1)\n        if latent_b.ndim == 3:\n            latent_b = latent_b.unsqueeze(0)\n\n        # ---- text ----\n        enc_b = text_cache[key[0]].to(device, dtype=dtype)  # [1, T, 7680]\n\n        if epoch == 1:\n            print(\"latent:\", latent_b.shape)   # [1,32,128,128]\n            print(\"text:\", enc_b.shape)        # [1,128,7680]\n\n        # ---- patchify ----\n        tokens = patchify_latents(latent_b)   # [1,4096,128]\n        tokens = torch.clamp(tokens, -CLAMP_VAL, CLAMP_VAL)\n\n        # ---- flow matching ----\n        eps = torch.randn_like(tokens)\n        eps = torch.clamp(eps, -CLAMP_VAL, CLAMP_VAL)\n\n        t = torch.rand(tokens.size(0), device=device, dtype=dtype)\n\n        z_t = (1 - t[:, None, None]) * eps + t[:, None, None] * tokens\n        target = tokens - eps\n        t_embed = t * FLOW_T_SCALE\n\n        # ---- pos ids ----\n        img_ids, txt_ids = generate_flux_pos_ids(\n            tokens.size(0), 64, 64, enc_b.size(1), device, dtype\n        )\n\n        # ---- forward ----\n        with torch.autocast(\"cuda\", dtype=torch.float16):\n            pred = pipe.transformer(\n                hidden_states=z_t,                 # NOT embedded\n                timestep=t_embed,\n                encoder_hidden_states=enc_b,\n                img_ids=img_ids,\n                txt_ids=txt_ids,\n                return_dict=False,\n            )[0]\n\n            loss = F.mse_loss(pred.float(), target.float())\n\n        # ---- backward ----\n        loss.backward()\n        torch.nn.utils.clip_grad_norm_(trainable_params, GRAD_CLIP)\n        optimizer.step()\n        optimizer.zero_grad(set_to_none=True)\n\n        epoch_loss += loss.item()\n\n    avg_loss = epoch_loss / max(1, len(loader))\n    print(f\"Epoch {epoch} | Avg Loss: {avg_loss:.6f}\")\n\n    # ---- checkpoint ----\n    if epoch % SAVE_EVERY == 0:\n        save_path = os.path.join(SAVE_DIR, f\"flux_klein_lora_epoch_{epoch:03d}.safetensors\")\n        lora_state = get_peft_model_state_dict(pipe.transformer)\n        save_file(lora_state, save_path)\n        print(\"💾 Saved LoRA:\", save_path)\n\n    if time.time() - start_time >= MAX_SECONDS:\n        break\n\n\n# ---------------- Final Save ----------------\n\nfinal_path = os.path.join(SAVE_DIR, \"flux_klein_lora_final.safetensors\")\nlora_state = get_peft_model_state_dict(pipe.transformer)\nsave_file(lora_state, final_path)\n\nprint(\"✅ Final FLUX LoRA saved:\", final_path)\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-21T06:59:02.591112Z","iopub.execute_input":"2026-01-21T06:59:02.591450Z","iopub.status.idle":"2026-01-21T07:46:34.817671Z","shell.execute_reply.started":"2026-01-21T06:59:02.591419Z","shell.execute_reply":"2026-01-21T07:46:34.816933Z"},"collapsed":true,"jupyter":{"outputs_hidden":true}},"outputs":[{"name":"stdout","text":"⏱️ Training started. Max time: 11.666666666666666 hours\n\n===== Epoch 1 =====\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   0%|          | 0/125 [00:00<?, ?it/s]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   1%|          | 1/125 [00:22<47:05, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   2%|▏         | 2/125 [00:45<46:29, 22.68s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   2%|▏         | 3/125 [01:08<46:05, 22.67s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   3%|▎         | 4/125 [01:30<45:41, 22.66s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   4%|▍         | 5/125 [01:53<45:19, 22.66s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   5%|▍         | 6/125 [02:16<45:00, 22.69s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   6%|▌         | 7/125 [02:38<44:40, 22.72s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   6%|▋         | 8/125 [03:01<44:20, 22.74s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   7%|▋         | 9/125 [03:24<43:59, 22.75s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   8%|▊         | 10/125 [03:47<43:37, 22.76s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   9%|▉         | 11/125 [04:10<43:17, 22.78s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  10%|▉         | 12/125 [04:32<42:55, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  10%|█         | 13/125 [04:55<42:31, 22.78s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  11%|█         | 14/125 [05:18<42:09, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  12%|█▏        | 15/125 [05:41<41:48, 22.81s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  13%|█▎        | 16/125 [06:04<41:25, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  14%|█▎        | 17/125 [06:26<41:01, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  14%|█▍        | 18/125 [06:49<40:39, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  15%|█▌        | 19/125 [07:12<40:16, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  16%|█▌        | 20/125 [07:35<39:52, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  17%|█▋        | 21/125 [07:58<39:29, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  18%|█▊        | 22/125 [08:20<39:07, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  18%|█▊        | 23/125 [08:43<38:45, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  19%|█▉        | 24/125 [09:06<38:21, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  20%|██        | 25/125 [09:29<37:58, 22.79s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  21%|██        | 26/125 [09:51<37:36, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  22%|██▏       | 27/125 [10:14<37:14, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  22%|██▏       | 28/125 [10:37<36:51, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  23%|██▎       | 29/125 [11:00<36:28, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  24%|██▍       | 30/125 [11:23<36:06, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  25%|██▍       | 31/125 [11:46<35:43, 22.81s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  26%|██▌       | 32/125 [12:08<35:21, 22.81s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  26%|██▋       | 33/125 [12:31<34:59, 22.82s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  27%|██▋       | 34/125 [12:54<34:37, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  28%|██▊       | 35/125 [13:17<34:15, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  29%|██▉       | 36/125 [13:40<33:51, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  30%|██▉       | 37/125 [14:03<33:29, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  30%|███       | 38/125 [14:25<33:05, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  31%|███       | 39/125 [14:48<32:42, 22.82s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  32%|███▏      | 40/125 [15:11<32:20, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  33%|███▎      | 41/125 [15:34<31:57, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  34%|███▎      | 42/125 [15:57<31:36, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  34%|███▍      | 43/125 [16:20<31:14, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  35%|███▌      | 44/125 [16:43<30:51, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  36%|███▌      | 45/125 [17:05<30:28, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  37%|███▋      | 46/125 [17:28<30:06, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  38%|███▊      | 47/125 [17:51<29:42, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  38%|███▊      | 48/125 [18:14<29:18, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  39%|███▉      | 49/125 [18:37<28:55, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  40%|████      | 50/125 [19:00<28:32, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  41%|████      | 51/125 [19:22<28:09, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  42%|████▏     | 52/125 [19:45<27:46, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  42%|████▏     | 53/125 [20:08<27:23, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  43%|████▎     | 54/125 [20:31<27:01, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  44%|████▍     | 55/125 [20:54<26:37, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  45%|████▍     | 56/125 [21:16<26:15, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  46%|████▌     | 57/125 [21:39<25:52, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  46%|████▋     | 58/125 [22:02<25:29, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  47%|████▋     | 59/125 [22:25<25:06, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  48%|████▊     | 60/125 [22:48<24:44, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  49%|████▉     | 61/125 [23:11<24:21, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  50%|████▉     | 62/125 [23:33<23:58, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  50%|█████     | 63/125 [23:56<23:36, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  51%|█████     | 64/125 [24:19<23:13, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  52%|█████▏    | 65/125 [24:42<22:51, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  53%|█████▎    | 66/125 [25:05<22:28, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  54%|█████▎    | 67/125 [25:28<22:05, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  54%|█████▍    | 68/125 [25:51<21:42, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  55%|█████▌    | 69/125 [26:13<21:19, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  56%|█████▌    | 70/125 [26:36<20:56, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  57%|█████▋    | 71/125 [26:59<20:34, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  58%|█████▊    | 72/125 [27:22<20:10, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  58%|█████▊    | 73/125 [27:45<19:48, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  59%|█████▉    | 74/125 [28:08<19:24, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  60%|██████    | 75/125 [28:31<19:01, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  61%|██████    | 76/125 [28:53<18:39, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  62%|██████▏   | 77/125 [29:16<18:17, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  62%|██████▏   | 78/125 [29:39<17:54, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  63%|██████▎   | 79/125 [30:02<17:31, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  64%|██████▍   | 80/125 [30:25<17:08, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  65%|██████▍   | 81/125 [30:48<16:45, 22.86s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  66%|██████▌   | 82/125 [31:11<16:22, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  66%|██████▋   | 83/125 [31:33<15:59, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  67%|██████▋   | 84/125 [31:56<15:36, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  68%|██████▊   | 85/125 [32:19<15:13, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  69%|██████▉   | 86/125 [32:42<14:50, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  70%|██████▉   | 87/125 [33:05<14:27, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  70%|███████   | 88/125 [33:28<14:04, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  71%|███████   | 89/125 [33:50<13:42, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  72%|███████▏  | 90/125 [34:13<13:19, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  73%|███████▎  | 91/125 [34:36<12:56, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  74%|███████▎  | 92/125 [34:59<12:33, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  74%|███████▍  | 93/125 [35:22<12:10, 22.82s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  75%|███████▌  | 94/125 [35:44<11:46, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  76%|███████▌  | 95/125 [36:07<11:23, 22.78s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  77%|███████▋  | 96/125 [36:30<11:00, 22.77s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  78%|███████▊  | 97/125 [36:53<10:37, 22.76s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  78%|███████▊  | 98/125 [37:15<10:14, 22.75s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  79%|███████▉  | 99/125 [37:38<09:51, 22.74s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  80%|████████  | 100/125 [38:01<09:28, 22.73s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  81%|████████  | 101/125 [38:24<09:05, 22.73s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  82%|████████▏ | 102/125 [38:46<08:42, 22.74s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  82%|████████▏ | 103/125 [39:09<08:20, 22.75s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  83%|████████▎ | 104/125 [39:32<07:58, 22.78s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  84%|████████▍ | 105/125 [39:55<07:36, 22.80s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  85%|████████▍ | 106/125 [40:18<07:13, 22.82s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  86%|████████▌ | 107/125 [40:41<06:50, 22.82s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  86%|████████▋ | 108/125 [41:03<06:28, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  87%|████████▋ | 109/125 [41:26<06:05, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  88%|████████▊ | 110/125 [41:49<05:42, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  89%|████████▉ | 111/125 [42:12<05:19, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  90%|████████▉ | 112/125 [42:35<04:56, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  90%|█████████ | 113/125 [42:58<04:34, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  91%|█████████ | 114/125 [43:20<04:11, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  92%|█████████▏| 115/125 [43:43<03:48, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  93%|█████████▎| 116/125 [44:06<03:25, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  94%|█████████▎| 117/125 [44:29<03:02, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  94%|█████████▍| 118/125 [44:52<02:39, 22.83s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  95%|█████████▌| 119/125 [45:15<02:17, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  96%|█████████▌| 120/125 [45:37<01:54, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  97%|█████████▋| 121/125 [46:00<01:31, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  98%|█████████▊| 122/125 [46:23<01:08, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  98%|█████████▊| 123/125 [46:46<00:45, 22.85s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:  99%|█████████▉| 124/125 [47:09<00:22, 22.84s/it]","output_type":"stream"},{"name":"stdout","text":"latent: torch.Size([1, 32, 128, 128])\ntext: torch.Size([1, 128, 7680])\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1: 100%|██████████| 125/125 [47:32<00:00, 22.82s/it]","output_type":"stream"},{"name":"stdout","text":"Epoch 1 | Avg Loss: 4.059623\n✅ Final FLUX LoRA saved: /kaggle/working/flux_klein_lora/flux_klein_lora_final.safetensors\n","output_type":"stream"},{"name":"stderr","text":"\n","output_type":"stream"}],"execution_count":17}]}
\ No newline at end of file
+{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.12.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":14582918,"sourceType":"datasetVersion","datasetId":8022630}],"dockerImageVersionId":31260,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"HOW TO USE:\n1) Upload a civitai dataset .zip file to your google drive named kaggleset.zip \n2) Use the https://huggingface.co/datasets/codeShare/lora-training-data/blob/main/civit_dataset_to_latent.ipynb notebook \nto convert this dataset to flux_captions.json and flux_latents.safetensors (saved to your drive upon running the script)\n3) Create a private dataset called image-caption-dataset\n4) Add the flux_captions.json and flux_latents.safetensor to this dataset\n5) In this notebook , press the '+ Add input' button and select your private dataset\n6) Run this notebook\n//----//\nIf you have ideas on improvements / developments on FLUX Klein 4B LoRa \ntraining let me know in the comment section of this repo","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# CELL 1 — Install libraries & force latest diffusers git (required for Flux2KleinPipeline)\n\nimport os\nos.environ[\"PYTORCH_CUDA_ALLOC_CONF\"] = \"expandable_segments:True\"\n\nprint(\"Uninstalling old packages...\")\n!pip uninstall -y torch torchvision torchaudio diffusers accelerate peft transformers bitsandbytes\n\nprint(\"Installing torch 2.5.1 cu121...\")\n!pip install --no-deps torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu121\n\nprint(\"Installing/upgrading other packages...\")\n!pip install --upgrade --no-cache-dir accelerate peft safetensors tqdm huggingface-hub\n!pip install --no-cache-dir transformers bitsandbytes\n\nprint(\"Installing latest diffusers from git (contains Flux2KleinPipeline)...\")\n!pip install -U git+https://github.com/huggingface/diffusers.git\n\nprint(\"\\nVERY IMPORTANT:\")\nprint(\"After this cell finishes → RESTART THE KERNEL / RUNTIME\")\nprint(\"Then run the remaining cells.\")\nprint(\"Without restart the new diffusers version will not be loaded.\")\n\nimport torch\nimport gc\nprint(\"Torch version:\", torch.__version__)\nprint(\"CUDA available:\", torch.cuda.is_available())\nprint(\"GPU:\", torch.cuda.get_device_name(0) if torch.cuda.is_available() else \"None\")\nprint(\"VRAM:\", round(torch.cuda.get_device_properties(0).total_memory / 1e9, 2), \"GB\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-22T20:32:58.996612Z","iopub.execute_input":"2026-01-22T20:32:58.997272Z"}},"outputs":[{"name":"stdout","text":"Uninstalling old packages...\nFound existing installation: torch 2.8.0+cu126\nUninstalling torch-2.8.0+cu126:\n  Successfully uninstalled torch-2.8.0+cu126\nFound existing installation: torchvision 0.23.0+cu126\nUninstalling torchvision-0.23.0+cu126:\n  Successfully uninstalled torchvision-0.23.0+cu126\nFound existing installation: torchaudio 2.8.0+cu126\nUninstalling torchaudio-2.8.0+cu126:\n  Successfully uninstalled torchaudio-2.8.0+cu126\nFound existing installation: diffusers 0.35.2\nUninstalling diffusers-0.35.2:\n  Successfully uninstalled diffusers-0.35.2\nFound existing installation: accelerate 1.11.0\nUninstalling accelerate-1.11.0:\n  Successfully uninstalled accelerate-1.11.0\nFound existing installation: peft 0.17.1\nUninstalling peft-0.17.1:\n  Successfully uninstalled peft-0.17.1\nFound existing installation: transformers 4.57.1\nUninstalling transformers-4.57.1:\n  Successfully uninstalled transformers-4.57.1\n\u001b[33mWARNING: Skipping bitsandbytes as it is not installed.\u001b[0m\u001b[33m\n\u001b[0m","output_type":"stream"}],"execution_count":null},{"cell_type":"code","source":"# CELL 2 — Config\n\nimport torch\nimport gc \nimport os\n\nMODEL_ID = \"black-forest-labs/FLUX.2-klein-4B\"\n\ndevice = \"cuda\"\ndtype  = torch.float16\n\nDATASET_NAME = \"image-caption-dataset\"\nCAPTIONS_PATH = f\"/kaggle/input/{DATASET_NAME}/flux_captions.json\"\nLATENTS_PATH  = f\"/kaggle/input/{DATASET_NAME}/flux_latents_klein.safetensors\"\n\nCACHE_DIR = \"/kaggle/working/cache\"\nSAVE_DIR  = \"/kaggle/working/flux_klein_lora\"\n\nos.makedirs(CACHE_DIR, exist_ok=True)\nos.makedirs(SAVE_DIR, exist_ok=True)\n\nBATCH_SIZE = 1\nLR         = 5e-5          # Reduced for stability\nRANK       = 4             # Reduced from 8\nALPHA      = 4             # Match rank\nEPOCHS     = 1\nSAVE_EVERY = 1\n\nFLOW_T_SCALE = 25.0        # Reduced from 50.0\nGRAD_CLIP    = 0.1         # Tighter clipping\nACCUM_STEPS  = 8           # Increased accumulation\n\nMAX_SECONDS = (11 * 60 + 40) * 60\n\nprint(f\"Config → rank={RANK}, alpha={ALPHA}, accum={ACCUM_STEPS}, batch={BATCH_SIZE}, lr={LR}, flow_scale={FLOW_T_SCALE}\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-22T19:19:23.783193Z","iopub.execute_input":"2026-01-22T19:19:23.783545Z","iopub.status.idle":"2026-01-22T19:19:25.533920Z","shell.execute_reply.started":"2026-01-22T19:19:23.783512Z","shell.execute_reply":"2026-01-22T19:19:25.533216Z"}},"outputs":[{"name":"stdout","text":"Config → rank=8, accum=4, batch=1\n","output_type":"stream"}],"execution_count":2},{"cell_type":"code","source":"# CELL 3 — Load captions dictionary\n\nimport json\n\nwith open(CAPTIONS_PATH, 'r', encoding='utf-8') as f:\n    captions_dict = json.load(f)\n\nprint(f\"Loaded {len(captions_dict)} captions\")\nprint(f\"First key: {list(captions_dict.keys())[0]}\")\nprint(f\"First caption (preview): {list(captions_dict.values())[0][:120]}...\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-22T19:19:27.800194Z","iopub.execute_input":"2026-01-22T19:19:27.801146Z","iopub.status.idle":"2026-01-22T19:19:27.815873Z","shell.execute_reply.started":"2026-01-22T19:19:27.801111Z","shell.execute_reply":"2026-01-22T19:19:27.815000Z"}},"outputs":[{"name":"stdout","text":"Loaded 125 captions\nFirst key: 008\nFirst caption (preview): The image shows two long hair, breasts, smile, multiple girls, brown hair, black hair, navel, holding, 2girls, standing,...\n","output_type":"stream"}],"execution_count":3},{"cell_type":"code","source":"# CELL 4 — Dataset & DataLoader (latents + key)\n\nfrom torch.utils.data import Dataset, DataLoader\nfrom safetensors.torch import load_file\n\nlatents_dict = load_file(LATENTS_PATH)\nprint(f\"Loaded {len(latents_dict)} latents\")\n\nkeys = list(latents_dict.keys())\n\nclass FluxTrainingDataset(Dataset):\n    def __init__(self, keys, latents_dict):\n        self.keys = keys\n        self.latents_dict = latents_dict\n\n    def __len__(self):\n        return len(self.keys)\n\n    def __getitem__(self, idx):\n        k = self.keys[idx]\n        latent = self.latents_dict[k]\n        if latent.dim() == 3:\n            latent = latent.unsqueeze(0)\n        return latent.to(dtype=torch.float16), k\n\ndataset = FluxTrainingDataset(keys, latents_dict)\n\ndataloader = DataLoader(\n    dataset,\n    batch_size=BATCH_SIZE,\n    shuffle=True,\n    num_workers=0,\n    pin_memory=True,\n    drop_last=True\n)\n\nprint(f\"DataLoader ready — {len(dataset)} samples, batch={BATCH_SIZE}\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-22T19:19:30.415451Z","iopub.execute_input":"2026-01-22T19:19:30.416463Z","iopub.status.idle":"2026-01-22T19:19:30.900882Z","shell.execute_reply.started":"2026-01-22T19:19:30.416415Z","shell.execute_reply":"2026-01-22T19:19:30.900087Z"}},"outputs":[{"name":"stdout","text":"Loaded 125 latents\nDataLoader ready — 125 samples, batch=1\n","output_type":"stream"}],"execution_count":4},{"cell_type":"code","source":"# CELL 5 — Load pipeline with CPU offload + LoRA\n\nfrom diffusers import Flux2KleinPipeline\nfrom peft import LoraConfig, get_peft_model\n\npipe = Flux2KleinPipeline.from_pretrained(\n    MODEL_ID,\n    dtype=dtype,\n    cache_dir=CACHE_DIR,\n)\n\npipe.enable_model_cpu_offload()\n\ntransformer = pipe.transformer.to(device, dtype=dtype)\n\nTARGET_MODULES = [\n    \"attn.to_q\", \"attn.to_k\", \"attn.to_v\", \"attn.to_out.0\",\n    \"attn.add_q_proj\", \"attn.add_k_proj\", \"attn.add_v_proj\", \"attn.to_add_out\",\n    \"ff.linear_in\", \"ff.linear_out\",\n    \"ff_context.linear_in\", \"ff_context.linear_out\",\n]\n\nlora_config = LoraConfig(\n    r=RANK,\n    lora_alpha=ALPHA,\n    target_modules=TARGET_MODULES,\n    lora_dropout=0.05,\n    bias=\"none\",\n)\n\ntransformer = get_peft_model(transformer, lora_config)\n\nprint(\"Transformer device:\", next(transformer.parameters()).device)\nprint(\"Text encoder device (idle):\", next(pipe.text_encoder.parameters()).device)\n\ntrainable = sum(p.numel() for p in transformer.parameters() if p.requires_grad)\ntotal = sum(p.numel() for p in transformer.parameters())\nprint(f\"Trainable: {trainable/1e6:.2f} M / Total: {total/1e6:.2f} M ({trainable/total*100:.3f} %)\")\n\n# Sanity check after LoRA\nprint(\"\\nVRAM after model + LoRA load:\")\nprint(torch.cuda.memory_summary(device=device, abbreviated=True))\n\ngc.collect()\ntorch.cuda.empty_cache()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-22T19:19:33.781103Z","iopub.execute_input":"2026-01-22T19:19:33.781782Z","iopub.status.idle":"2026-01-22T19:22:56.924440Z","shell.execute_reply.started":"2026-01-22T19:19:33.781735Z","shell.execute_reply":"2026-01-22T19:22:56.923479Z"}},"outputs":[{"name":"stderr","text":"2026-01-22 19:19:44.784104: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\nWARNING: All log messages before absl::InitializeLog() is called are written to STDERR\nE0000 00:00:1769109584.993343     186 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\nE0000 00:00:1769109585.053346     186 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\nW0000 00:00:1769109585.557981     186 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\nW0000 00:00:1769109585.558032     186 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\nW0000 00:00:1769109585.558035     186 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\nW0000 00:00:1769109585.558038     186 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\nFlax classes are deprecated and will be removed in Diffusers v1.0.0. We recommend migrating to PyTorch classes or pinning your version of Diffusers.\nFlax classes are deprecated and will be removed in Diffusers v1.0.0. We recommend migrating to PyTorch classes or pinning your version of Diffusers.\n","output_type":"stream"},{"output_type":"display_data","data":{"text/plain":"model_index.json:   0%|          | 0.00/446 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"e5ccc8533dcb459ea81c3752b77dedfe"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Fetching 17 files:   0%|          | 0/17 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"38725d1dd178407da911dd346494fa80"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"merges.txt: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"3de4984338b54f81bb5e3ea34dabd1e2"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"chat_template.jinja: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"02bbf2b0a9ef41abb8dbed58ea812428"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"config.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"eb2dc68f547b45f8829d7667b24bae4e"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"model.safetensors.index.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"3ed7b61c2251437eb888a5f49563139a"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"added_tokens.json:   0%|          | 0.00/707 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"dae22926b8964b7a8479c8cf8685ffbe"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"text_encoder/model-00001-of-00002.safete(…):   0%|          | 0.00/4.97G [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"61908f03443d4339a9148d8c71a82361"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"scheduler_config.json:   0%|          | 0.00/486 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"a27aba830a0448ca82bd37f2802d725c"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"text_encoder/model-00002-of-00002.safete(…):   0%|          | 0.00/3.08G [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"561a71e583e145d69a320b60cd62a0d1"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"special_tokens_map.json:   0%|          | 0.00/613 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"17a166f3be8242aa866211470afba535"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"vocab.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"ff4378d1960a485a9a7d72091449bf89"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"tokenizer/tokenizer.json:   0%|          | 0.00/11.4M [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"7c24d3f8b06f46bcaa6079f0b86e679c"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"tokenizer_config.json: 0.00B [00:00, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"b81c35acb4c64b5796bae47bacce893e"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"config.json:   0%|          | 0.00/541 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"0932ea6388cc469b9cb02405e51924db"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"transformer/diffusion_pytorch_model.safe(…):   0%|          | 0.00/7.75G [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"e836c9e259ed4a1f98239c2bad5c77b1"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"config.json:   0%|          | 0.00/821 [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"7dbb08adcbf241888a7aa374ce61f414"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"vae/diffusion_pytorch_model.safetensors:   0%|          | 0.00/168M [00:00<?, ?B/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"b83d1a1e20f545d2846c9d1dbad61bed"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Loading pipeline components...:   0%|          | 0/5 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"1631c6a679f24598a12cb46dc594521f"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"d7b28b3bdcfb40629c1684ab9d2c9dc3"}},"metadata":{}},{"name":"stdout","text":"Transformer device: cuda:0\nText encoder device (idle): cpu\nTrainable: 4.67 M / Total: 3880.21 M (0.120 %)\n\nVRAM after setup:\n|===========================================================================|\n|                  PyTorch CUDA memory summary, device ID 0                 |\n|---------------------------------------------------------------------------|\n|            CUDA OOMs: 0            |        cudaMalloc retries: 0         |\n|===========================================================================|\n|        Metric         | Cur Usage  | Peak Usage | Tot Alloc  | Tot Freed  |\n|---------------------------------------------------------------------------|\n| Allocated memory      |   7411 MiB |   7411 MiB |   7420 MiB |   9120 KiB |\n|---------------------------------------------------------------------------|\n| Active memory         |   7411 MiB |   7411 MiB |   7420 MiB |   9120 KiB |\n|---------------------------------------------------------------------------|\n| Requested memory      |   7409 MiB |   7409 MiB |   7418 MiB |   9120 KiB |\n|---------------------------------------------------------------------------|\n| GPU reserved memory   |   7412 MiB |   7412 MiB |   7412 MiB |      0 B   |\n|---------------------------------------------------------------------------|\n| Non-releasable memory | 690176 B   |  18944 KiB |  45232 KiB |  44558 KiB |\n|---------------------------------------------------------------------------|\n| Allocations           |     289    |     290    |     409    |     120    |\n|---------------------------------------------------------------------------|\n| Active allocs         |     289    |     290    |     409    |     120    |\n|---------------------------------------------------------------------------|\n| GPU reserved segments |     116    |     116    |     116    |       0    |\n|---------------------------------------------------------------------------|\n| Non-releasable allocs |      11    |      11    |      56    |      45    |\n|---------------------------------------------------------------------------|\n| Oversize allocations  |       0    |       0    |       0    |       0    |\n|---------------------------------------------------------------------------|\n| Oversize GPU segments |       0    |       0    |       0    |       0    |\n|===========================================================================|\n\n","output_type":"stream"}],"execution_count":5},{"cell_type":"code","source":"# CELL 6 — Training loop with gradient accumulation & on-the-fly encoding (optimized)\n\nfrom bitsandbytes.optim import AdamW8bit\nfrom torch.optim.lr_scheduler import CosineAnnealingLR\nfrom tqdm import tqdm\nimport time\nimport gc\nimport torch.nn.functional as F\n\noptimizer = AdamW8bit(transformer.parameters(), lr=LR, weight_decay=0.01)\n\nlr_scheduler = CosineAnnealingLR(\n    optimizer,\n    T_max=len(dataloader) * EPOCHS * ACCUM_STEPS,\n    eta_min=LR * 0.1\n)\n\n# Patch positions (4096 image tokens, 4 coords)\nPATCH_H = PATCH_W = 64\nys, xs = torch.meshgrid(\n    torch.arange(PATCH_H, device=device),\n    torch.arange(PATCH_W, device=device),\n    indexing=\"ij\"\n)\n\nimg_pos_ids = torch.stack(\n    [\n        xs.flatten(),\n        ys.flatten(),\n        torch.zeros_like(xs.flatten()),\n        torch.zeros_like(xs.flatten()),\n    ],\n    dim=-1,\n).unsqueeze(0)  # (1, 4096, 4)\n\nprint(\"img_pos_ids shape:\", img_pos_ids.shape)\n\ntransformer.train()\n\nstart_time = time.time()\nglobal_step = 0\n\nprint(f\"\\nStarting training — effective batch={BATCH_SIZE * ACCUM_STEPS}, rank={RANK}\")\n\nfor epoch in range(EPOCHS):\n    print(f\"\\n===== Epoch {epoch+1}/{EPOCHS} =====\")\n    print(f\"Epoch start LR: {optimizer.param_groups[0]['lr']:.2e}\")\n    \n    progress = tqdm(dataloader, desc=f\"Epoch {epoch+1}\")\n\n    accum_counter = 0\n\n    for batch_idx, (latent, key) in enumerate(progress):\n\n        if time.time() - start_time > MAX_SECONDS:\n            print(\"Time limit reached.\")\n            break\n\n        latent = latent.to(device, non_blocking=True).contiguous()\n        if latent.dim() == 5 and latent.shape[1] == 1:\n            latent = latent.squeeze(1)\n\n        if isinstance(key, (list, tuple)):\n            key = key[0]\n\n        caption = captions_dict[key]\n\n        inputs = pipe.tokenizer(\n            caption,\n            return_tensors=\"pt\",\n            padding=\"longest\",\n            truncation=True,\n            max_length=256,\n            return_attention_mask=True,\n        ).to(device)\n\n        with torch.no_grad():\n            outputs = pipe.text_encoder(**inputs, output_hidden_states=True)\n\n        hs = outputs.hidden_states\n        encoder_hidden_states = torch.cat([hs[9], hs[18], hs[27]], dim=-1).to(dtype).contiguous()\n\n        del outputs, inputs, hs\n        gc.collect()\n        torch.cuda.empty_cache()\n\n        # ---- latent -> patches ----\n        B, C, H_lat, W_lat = latent.shape  # expect (B, 32, 128, 128)\n\n        latent_patched = latent.reshape(B, 32, 64, 2, 64, 2)\n        latent_patched = latent_patched.permute(0, 2, 4, 1, 3, 5)\n        latent_patched = latent_patched.reshape(B, 4096, 128).contiguous()\n\n        noise = torch.randn_like(latent_patched)\n        t = torch.rand(B, device=device)\n        noisy_latent_patched = latent_patched + noise * t.view(B, 1, 1) * FLOW_T_SCALE\n\n        # Sanity check: noisy latent\n        print(f\"Noisy latent shape: {noisy_latent_patched.shape}, dtype: {noisy_latent_patched.dtype}\")\n        assert noisy_latent_patched.shape == (B, 4096, 128), \"Unexpected noisy latent shape!\"\n\n        # ---- text position ids ----\n        seq_len = encoder_hidden_states.shape[1]\n        txt_pos_ids = torch.zeros((B, seq_len, 4), device=device)\n        txt_pos_ids[..., 2] = torch.arange(seq_len, device=device).unsqueeze(0)\n\n        # ---- image position ids (expand to batch) ----\n        img_ids = img_pos_ids.expand(B, -1, -1)\n\n        # VRAM check before forward\n        vram_before = torch.cuda.memory_allocated() / 1e9\n        print(f\"VRAM before forward: {vram_before:.2f} GB\")\n\n        with torch.autocast(\"cuda\", dtype=dtype):\n            pred = transformer(\n                hidden_states=noisy_latent_patched,\n                encoder_hidden_states=encoder_hidden_states,\n                timestep=t,\n                img_ids=img_ids,\n                txt_ids=txt_pos_ids,\n                guidance=None,\n                return_dict=False,\n            )[0]\n\n            loss = F.mse_loss(pred.float(), noise.float())\n\n        loss = loss / ACCUM_STEPS\n        loss.backward()\n\n        accum_counter += 1\n        if accum_counter % ACCUM_STEPS == 0:\n            grad_norm = torch.nn.utils.clip_grad_norm_(transformer.parameters(), GRAD_CLIP)\n            \n            # Sanity: warn on high grad norm\n            if grad_norm.item() > 1.0:\n                print(f\"Warning: High grad norm ({grad_norm.item():.3f}) → consider lowering LR or increasing clip!\")\n            \n            optimizer.step()\n            optimizer.zero_grad(set_to_none=True)\n            lr_scheduler.step()\n        else:\n            grad_norm = torch.tensor(0.0, device=device)\n\n        global_step += 1\n\n        if global_step % 10 == 0 or batch_idx < 10:\n            current_lr = optimizer.param_groups[0][\"lr\"]\n            progress.set_postfix(\n                loss=f\"{loss.item() * ACCUM_STEPS:.5f}\",\n                lr=f\"{current_lr:.2e}\",\n                gnorm=f\"{grad_norm.item():.3f}\",\n                vram=f\"{torch.cuda.memory_allocated()/1e9:.1f}GB\",\n            )\n\n        # Extra clear after each step\n        del pred, loss, noisy_latent_patched, noise, latent_patched, encoder_hidden_states\n        gc.collect()\n        torch.cuda.empty_cache()\n\n        # Sanity: VRAM after clear\n        vram_after = torch.cuda.memory_allocated() / 1e9\n        if global_step % 5 == 0:\n            print(f\"VRAM after clear (step {global_step}): {vram_after:.2f} GB (freed: {vram_before - vram_after:.2f} GB)\")\n\n    print(f\"Epoch {epoch+1} end LR: {optimizer.param_groups[0]['lr']:.2e}\")\n\n    if (epoch + 1) % SAVE_EVERY == 0 or epoch == EPOCHS - 1:\n        save_path = f\"{SAVE_DIR}/flux-klein-lora_r{RANK}_e{epoch+1}_step{global_step}.safetensors\"\n        from safetensors.torch import save_file\n        save_file(get_peft_model_state_dict(transformer), save_path)\n        print(f\"Saved: {os.path.basename(save_path)}\")\n\nprint(\"\\nTraining finished.\")\nprint(f\"Total steps: {global_step}\")\nprint(f\"Elapsed: {(time.time() - start_time)/3600:.2f} hours\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-22T19:26:08.317120Z","iopub.execute_input":"2026-01-22T19:26:08.317501Z","iopub.status.idle":"2026-01-22T19:26:19.989518Z","shell.execute_reply.started":"2026-01-22T19:26:08.317473Z","shell.execute_reply":"2026-01-22T19:26:19.988121Z"}},"outputs":[{"name":"stdout","text":"img_pos_ids shape: torch.Size([1, 4096, 4])\n\nStarting training — effective batch=4, rank=8\n\n===== Epoch 1/1 =====\n","output_type":"stream"},{"name":"stderr","text":"Epoch 1:   0%|          | 0/125 [00:11<?, ?it/s]\n","output_type":"stream"},{"traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mOutOfMemoryError\u001b[0m                          Traceback (most recent call last)","\u001b[0;32m/tmp/ipykernel_186/4095702193.py\u001b[0m in \u001b[0;36m<cell line: 0>\u001b[0;34m()\u001b[0m\n\u001b[1;32m    106\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    107\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mautocast\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"cuda\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 108\u001b[0;31m             pred = transformer(\n\u001b[0m\u001b[1;32m    109\u001b[0m                 \u001b[0mhidden_states\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnoisy_latent_patched\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    110\u001b[0m                 \u001b[0mencoder_hidden_states\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mencoder_hidden_states\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1734\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_compiled_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# type: ignore[misc]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1735\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1736\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1737\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1738\u001b[0m     \u001b[0;31m# torchrec tests the code consistency with the following code\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1745\u001b[0m                 \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_pre_hooks\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_hooks\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1746\u001b[0m                 or _global_forward_hooks or _global_forward_pre_hooks):\n\u001b[0;32m-> 1747\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mforward_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1748\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1749\u001b[0m         \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/peft/peft_model.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m    919\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_enable_peft_forward_hooks\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    920\u001b[0m             \u001b[0mkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mv\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mk\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mv\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mk\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mspecial_peft_forward_args\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 921\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_base_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    922\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    923\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mgenerate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1734\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_compiled_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# type: ignore[misc]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1735\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1736\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1737\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1738\u001b[0m     \u001b[0;31m# torchrec tests the code consistency with the following code\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1745\u001b[0m                 \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_pre_hooks\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_hooks\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1746\u001b[0m                 or _global_forward_hooks or _global_forward_pre_hooks):\n\u001b[0;32m-> 1747\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mforward_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1748\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1749\u001b[0m         \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/accelerate/hooks.py\u001b[0m in \u001b[0;36mnew_forward\u001b[0;34m(module, *args, **kwargs)\u001b[0m\n\u001b[1;32m    173\u001b[0m                 \u001b[0moutput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_old_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    174\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 175\u001b[0;31m             \u001b[0moutput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_old_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    176\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_hf_hook\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpost_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodule\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    177\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/diffusers/models/transformers/transformer_flux2.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, hidden_states, encoder_hidden_states, timestep, img_ids, txt_ids, guidance, joint_attention_kwargs, return_dict)\u001b[0m\n\u001b[1;32m    895\u001b[0m                 )\n\u001b[1;32m    896\u001b[0m             \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 897\u001b[0;31m                 hidden_states = block(\n\u001b[0m\u001b[1;32m    898\u001b[0m                     \u001b[0mhidden_states\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mhidden_states\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    899\u001b[0m                     \u001b[0mencoder_hidden_states\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1734\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_compiled_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# type: ignore[misc]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1735\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1736\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1737\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1738\u001b[0m     \u001b[0;31m# torchrec tests the code consistency with the following code\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1745\u001b[0m                 \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_pre_hooks\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_hooks\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1746\u001b[0m                 or _global_forward_hooks or _global_forward_pre_hooks):\n\u001b[0;32m-> 1747\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mforward_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1748\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1749\u001b[0m         \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/diffusers/models/transformers/transformer_flux2.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, hidden_states, encoder_hidden_states, temb_mod_params, image_rotary_emb, joint_attention_kwargs, split_hidden_states, text_seq_len)\u001b[0m\n\u001b[1;32m    443\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    444\u001b[0m         \u001b[0mjoint_attention_kwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mjoint_attention_kwargs\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 445\u001b[0;31m         attn_output = self.attn(\n\u001b[0m\u001b[1;32m    446\u001b[0m             \u001b[0mhidden_states\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnorm_hidden_states\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    447\u001b[0m             \u001b[0mimage_rotary_emb\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mimage_rotary_emb\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1734\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_compiled_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# type: ignore[misc]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1735\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1736\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call_impl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1737\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1738\u001b[0m     \u001b[0;31m# torchrec tests the code consistency with the following code\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1745\u001b[0m                 \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_pre_hooks\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0m_global_backward_hooks\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1746\u001b[0m                 or _global_forward_hooks or _global_forward_pre_hooks):\n\u001b[0;32m-> 1747\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mforward_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1748\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1749\u001b[0m         \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/diffusers/models/transformers/transformer_flux2.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, hidden_states, attention_mask, image_rotary_emb, **kwargs)\u001b[0m\n\u001b[1;32m    388\u001b[0m             )\n\u001b[1;32m    389\u001b[0m         \u001b[0mkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mw\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mk\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mw\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mk\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mattn_parameters\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 390\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprocessor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhidden_states\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mattention_mask\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mimage_rotary_emb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    391\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    392\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/diffusers/models/transformers/transformer_flux2.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, attn, hidden_states, attention_mask, image_rotary_emb)\u001b[0m\n\u001b[1;32m    307\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    308\u001b[0m         \u001b[0;31m# Concatenate and parallel output projection\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 309\u001b[0;31m         \u001b[0mhidden_states\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mhidden_states\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmlp_hidden_states\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    310\u001b[0m         \u001b[0mhidden_states\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mattn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto_out\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhidden_states\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    311\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mOutOfMemoryError\u001b[0m: CUDA out of memory. Tried to allocate 204.00 MiB. GPU 0 has a total capacity of 15.89 GiB of which 83.12 MiB is free. Process 5047 has 15.80 GiB memory in use. Of the allocated memory 15.15 GiB is allocated by PyTorch, and 377.02 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation.  See documentation for Memory Management  (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)"],"ename":"OutOfMemoryError","evalue":"CUDA out of memory. Tried to allocate 204.00 MiB. GPU 0 has a total capacity of 15.89 GiB of which 83.12 MiB is free. Process 5047 has 15.80 GiB memory in use. Of the allocated memory 15.15 GiB is allocated by PyTorch, and 377.02 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation.  See documentation for Memory Management  (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)","output_type":"error"}],"execution_count":9}]}
\ No newline at end of file