Code Improvement Suggestions

[18570426220]

This is a great project! However, I encountered the following issues while attempting to train it:
1.There is a discrepancy in the task argument between the train and play commands in DeepMimic, which causes errors when actually running play.
2.The AMP policy fails to train properly.

[18570426220]

I am a newcomer to the field of humanoid robotics. I want to start by training a model that can walk stably in real-world deployment. Could you give me some advice? Which one performs better, DeepMimic or AMP? Also, are the datasets provided in the current project sufficient for walking training?