RuntimeError: Error(s) in loading state_dict for ModuleDict:

[jisenlin]

Hello, author. While reproducing the Segmenting Human Demonstrations code, I encountered the following error:
Missing key(s) in state_dict: "policy.nets.decoder.nets.action.weight", "policy.nets.decoder.nets.action.bias".
Unexpected key(s) in state_dict: "policy.nets.decoder.nets.logit.weight", "policy.nets.decoder.nets.logit.bias".
I did some investigation and found that the IDM algorithm is instantiated in two different places: one is in the prime directory, and the other is in the robomimic directory. Why are there two separate algo implementations? What is the reason for this design?