⚙️ Running in WANDB offline mode ⚙️ Running in WANDB offline mode ⚙️ Running in WANDB offline mode ⚙️ Running in WANDB offline mode `torch_dtype` is deprecated! Use `dtype` instead! `torch_dtype` is deprecated! Use `dtype` instead! Loading vision encoder from pretrained_models/videollama3_7b_local ... `torch_dtype` is deprecated! Use `dtype` instead! `torch_dtype` is deprecated! Use `dtype` instead! `torch_dtype` is deprecated! Use `dtype` instead! `torch_dtype` is deprecated! Use `dtype` instead! `torch_dtype` is deprecated! Use `dtype` instead! `torch_dtype` is deprecated! Use `dtype` instead! Vision encoder loaded. hidden_size=1152 Compressor type=local_attn_conv layers=8 heads=8 HW=16×16 Fixed frames T=8 (decoder depth=log2(T)=3) SGM mask_ratio=0.75 recon_masked_only=True mask_temperature=1.0 The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. Force set image size to be 448 Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/ Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/ Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/ Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/ Dataset size: 977344 [2026-08-20 03:46:25,591] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2026-08-20 03:46:25,601] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2026-08-20 03:46:25,620] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2026-08-20 03:46:25,688] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect) 0%| | 0/19090 [00:00