diff --git "a/transfer/code/CCFM/logs/ccfm_v2_cached_5405774.out" "b/transfer/code/CCFM/logs/ccfm_v2_cached_5405774.out" new file mode 100644--- /dev/null +++ "b/transfer/code/CCFM/logs/ccfm_v2_cached_5405774.out" @@ -0,0 +1,689 @@ +========================================== +Job ID: 5405774 +Job Name: ccfm_v2_cached +Start: Sun Mar 15 16:58:15 JST 2026 +Node: b0034 +GPU: NVIDIA H100, 95830 MiB +Run: CCFM v2 cached (KV fix + loss fix + EMA + RK4 + logit-normal + warmup + scGPT cache) +========================================== +The following values were not passed to `accelerate launch` and had defaults used instead: + `--num_machines` was set to a value of `1` + `--mixed_precision` was set to a value of `'no'` + `--dynamo_backend` was set to a value of `'no'` +To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`. +/home/hp250092/ku50001222/qian/aivc/lfj/transfer/code/scGPT/scgpt/model/model.py:21: UserWarning: flash_attn is not installed + warnings.warn("flash_attn is not installed") +WARNING:accelerate.utils.other:[RANK 0] Detected kernel version 4.18.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher. +/home/hp250092/ku50001222/qian/aivc/lfj/transfer/code/scDFM/src/data_process/data.py:185: ImplicitModificationWarning: Trying to modify attribute `.obs` of view, initializing view as actual. + self.adata.obs['condition'] = self.adata.obs['condition'].str.replace('ctrl', 'control') +/home/hp250092/ku50001222/qian/aivc/lfj/stack_env/lib/python3.13/site-packages/scanpy/preprocessing/_highly_variable_genes.py:806: ImplicitModificationWarning: Trying to modify attribute `._uns` of view, initializing view as actual. + adata.uns["hvg"] = {"flavor": flavor} +/home/hp250092/ku50001222/qian/aivc/lfj/transfer/code/scDFM/src/data_process/data.py:277: ImplicitModificationWarning: Trying to modify attribute `.obs` of view, initializing view as actual. + self.adata.obs['perturbation_covariates'] = self.adata.obs[perturbation_covariates].apply(lambda x: '+'.join(x), axis=1) +/home/hp250092/ku50001222/qian/aivc/lfj/transfer/code/scDFM/src/data_process/data.py:329: ImplicitModificationWarning: Trying to modify attribute `.obs` of view, initializing view as actual. + self.adata.obs['perturbation_covariates'] = self.adata.obs[perturbation_covariates].apply(lambda x: '+'.join(x), axis=1) +/home/hp250092/ku50001222/qian/aivc/lfj/transfer/code/CCFM/src/data/scgpt_extractor.py:109: UserWarning: FrozenScGPTExtractor: 498/5035 HVG genes not found in scGPT vocab, will use zero vectors. + warnings.warn( +CascadedFlowConfig(model_type='cascaded', batch_size=48, ntoken=512, d_model=128, nhead=8, nlayers=4, lr=5e-05, steps=200000, eta_min=1e-06, devices='1', test_only=False, data_name='norman', perturbation_function='crisper', noise_type='Gaussian', poisson_alpha=0.8, poisson_target_sum=-1, print_every=10000, mode='predict_y', result_path='./result', fusion_method='differential_perceiver', infer_top_gene=1000, n_top_genes=5000, checkpoint_path='', gamma=0.5, split_method='additive', use_mmd_loss=True, fold=1, use_negative_edge=True, topk=30, scgpt_dim=512, bottleneck_dim=128, latent_weight=1.0, choose_latent_p=0.4, target_std=1.0, dh_depth=2, warmup_batches=200, ema_decay=0.9999, t_sample_mode='logit_normal', t_expr_mean=0.0, t_expr_std=1.0, t_latent_mean=0.0, t_latent_std=1.0, warmup_steps=2000, scgpt_model_dir='transfer/data/scGPT_pretrained', scgpt_max_seq_len=1200, scgpt_cache_path='scgpt_cache_norman.h5', latent_steps=20, expr_steps=20, ode_method='rk4') +Converted var_names to gene symbols, sample: ['RP11-34P13.8', 'RP11-54O7.3', 'SAMD11', 'PERM1', 'HES4'] +Warning: ctrl is not in the gene names +##### loading vocab from file ##### +Loaded 159/163 pretrained parameters +Using pre-extracted scGPT cache: scgpt_cache_norman.h5 + Cache shape: (91205, 5035, 512), cells: 91205 + 0%| | 0/200000 [00:00