[2026/09/28 20:29:01] hftrainer INFO: Config: configs/public_data/minimax_h3.py [2026/09/28 20:29:01] hftrainer INFO: Work dir: work_dirs/public_data/minimax_h3 [2026/09/28 20:29:02] hftrainer INFO: Config saved to: work_dirs/public_data/minimax_h3/20260928_202902/config.py [2026/09/28 20:29:02] hftrainer INFO: Work dir: work_dirs/public_data/minimax_h3 [2026/09/28 20:29:02] hftrainer INFO: Run dir (logs/tb): work_dirs/public_data/minimax_h3/20260928_202902 [2026/09/28 20:29:02] hftrainer INFO: Environment info: Platform: Linux-6.1.158-178.288.amzn2023.x86_64-x86_64-with-glibc2.39 Python: 3.12.3 (main, Nov 6 2024, 18:32:19) [GCC 13.2.0] PyTorch: 2.8.0+cu128 CUDA available: True CUDA version: 12.8 GPU count: 1 GPU 0: NVIDIA H200 accelerate: 1.15.0 mmengine: 0.10.7 safetensors: 0.8.0 torchvision: 0.23.0+cu128 datasets: not installed [2026/09/28 20:32:59] hftrainer INFO: Model Summary: Module Total Params Trainable Params Trainable Save Ckpt ----------- -------------- ---------------- --------- --------- transformer 33,128,211,200 5,218,304 True True ----------- -------------- ---------------- --------- --------- TOTAL 33,128,211,200 5,218,304 Trainable ratio: 0.02% [2026/09/28 20:33:12] hftrainer INFO: step [1/20] lr=1.00e-04 loss=0.7431 loss_video=0.3840 loss_audio=0.3591 video_timestep=0.8259 audio_timestep=0.7296 data_time=0.03s train_time=1.02s eta=0:00:19 [2026/09/28 20:33:13] hftrainer INFO: step [2/20] lr=1.00e-04 loss=0.8247 loss_video=0.4912 loss_audio=0.3335 video_timestep=0.2255 audio_timestep=0.3584 data_time=0.04s train_time=0.38s eta=0:00:13 [2026/09/28 20:33:13] hftrainer INFO: step [3/20] lr=1.00e-04 loss=0.5613 loss_video=0.2388 loss_audio=0.3225 video_timestep=0.5246 audio_timestep=0.3273 data_time=0.04s train_time=0.37s eta=0:00:10 [2026/09/28 20:33:13] hftrainer INFO: step [4/20] lr=1.00e-04 loss=0.8593 loss_video=0.4852 loss_audio=0.3742 video_timestep=0.2590 audio_timestep=0.7428 data_time=0.05s train_time=0.36s eta=0:00:09 [2026/09/28 20:33:14] hftrainer INFO: step [5/20] lr=1.00e-04 loss=0.6254 loss_video=0.2784 loss_audio=0.3470 video_timestep=0.6813 audio_timestep=0.3281 data_time=0.06s train_time=0.36s eta=0:00:08 [2026/09/28 20:33:14] hftrainer INFO: step [6/20] lr=1.00e-04 loss=0.9048 loss_video=0.6041 loss_audio=0.3007 video_timestep=0.9150 audio_timestep=0.2950 data_time=0.05s train_time=0.38s eta=0:00:07 [2026/09/28 20:33:15] hftrainer INFO: step [7/20] lr=1.00e-04 loss=0.9372 loss_video=0.5992 loss_audio=0.3380 video_timestep=0.1350 audio_timestep=0.6967 data_time=0.05s train_time=0.36s eta=0:00:06 [2026/09/28 20:33:15] hftrainer INFO: step [8/20] lr=1.00e-04 loss=0.6392 loss_video=0.2526 loss_audio=0.3866 video_timestep=0.5209 audio_timestep=0.4460 data_time=0.04s train_time=0.48s eta=0:00:06 [2026/09/28 20:33:16] hftrainer INFO: step [9/20] lr=1.00e-04 loss=0.6392 loss_video=0.2645 loss_audio=0.3747 video_timestep=0.5776 audio_timestep=0.5251 data_time=0.04s train_time=0.37s eta=0:00:05 [2026/09/28 20:33:16] hftrainer INFO: step [10/20] lr=1.00e-04 loss=0.6050 loss_video=0.2855 loss_audio=0.3195 video_timestep=0.6792 audio_timestep=0.2778 data_time=0.05s train_time=0.37s eta=0:00:04 [2026/09/28 20:33:17] hftrainer INFO: step [11/20] lr=1.00e-04 loss=0.9410 loss_video=0.5748 loss_audio=0.3662 video_timestep=0.1739 audio_timestep=0.6440 data_time=0.05s train_time=0.37s eta=0:00:04 [2026/09/28 20:33:17] hftrainer INFO: step [12/20] lr=1.00e-04 loss=0.6428 loss_video=0.4285 loss_audio=0.2143 video_timestep=0.3011 audio_timestep=0.0803 data_time=0.05s train_time=0.38s eta=0:00:03 [2026/09/28 20:33:17] hftrainer INFO: step [13/20] lr=1.00e-04 loss=0.9108 loss_video=0.5464 loss_audio=0.3644 video_timestep=0.8908 audio_timestep=0.7894 data_time=0.05s train_time=0.39s eta=0:00:03 [2026/09/28 20:33:18] hftrainer INFO: step [14/20] lr=1.00e-04 loss=0.6750 loss_video=0.3034 loss_audio=0.3717 video_timestep=0.7290 audio_timestep=0.7934 data_time=0.05s train_time=0.38s eta=0:00:02 [2026/09/28 20:33:18] hftrainer INFO: step [15/20] lr=1.00e-04 loss=0.6001 loss_video=0.3243 loss_audio=0.2758 video_timestep=0.7546 audio_timestep=0.1017 data_time=0.06s train_time=0.38s eta=0:00:02 [2026/09/28 20:33:19] hftrainer INFO: step [16/20] lr=1.00e-04 loss=0.7366 loss_video=0.3572 loss_audio=0.3795 video_timestep=0.7740 audio_timestep=0.4574 data_time=0.08s train_time=0.49s eta=0:00:01 [2026/09/28 20:33:19] hftrainer INFO: step [17/20] lr=1.00e-04 loss=0.6837 loss_video=0.3339 loss_audio=0.3498 video_timestep=0.7304 audio_timestep=0.7293 data_time=0.05s train_time=0.37s eta=0:00:01 [2026/09/28 20:33:20] hftrainer INFO: step [18/20] lr=1.00e-04 loss=0.6798 loss_video=0.3035 loss_audio=0.3763 video_timestep=0.4618 audio_timestep=0.8013 data_time=0.05s train_time=0.37s eta=0:00:00 [2026/09/28 20:33:20] hftrainer INFO: step [19/20] lr=1.00e-04 loss=0.7431 loss_video=0.4033 loss_audio=0.3398 video_timestep=0.2710 audio_timestep=0.7458 data_time=0.05s train_time=0.37s eta=0:00:00 [2026/09/28 20:33:20] hftrainer INFO: step [20/20] lr=1.00e-04 loss=0.6399 loss_video=0.3399 loss_audio=0.3000 video_timestep=0.7573 audio_timestep=0.3413 data_time=0.04s train_time=0.36s eta=0:00:00 [2026/09/28 20:33:21] hftrainer INFO: Saved checkpoint to: work_dirs/public_data/minimax_h3/checkpoint-iter_20 [2026/09/28 20:33:22] hftrainer INFO: step=20 val/loss=0.8745 val/loss_video=0.4547 val/loss_audio=0.4198 [2026/09/28 20:33:22] hftrainer INFO: Training complete.