[Train] Support Fastwam training

[Musisoul]

No description provided.

[gemini-code-assist]

Code Review

This pull request introduces the FastWAM (Mixture of Transformers) training pipeline, adding support for the LIBERO dataset, specialized preprocessing, normalization, and preparation assets. It implements the FastWAM model wrapper, ActionDiT expert, MoT architecture, and FastWAMTrainer to handle joint video and action training. The review feedback highlights several critical issues: an AttributeError in video_decoder.py caused by calling reader.container.close() on a VideoReader object; silent data corruption in processor.py due to in-place modification of a cached action tensor; potential device/dtype mismatches in normalizer.py when performing operations with CPU-bound scale and offset tensors; and a memory leak in lerobot_dataset.py caused by using @lru_cache on an instance method.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.