takuseno/d3rlpy

[QUESTION] How to add an action mask to bc algorithm?

Lucien-Evans-123 opened this issue · 2 comments

I want to add an action mask to bc algorithm compare to the masked cql. So could you please tell me where should I modify the code?
Thank you!

oh, it's discrete bc algorithm.

Oh, I'm sorry to bother you. I find where to modify! Thank you!
It's in bc_impl.py
def inner_predict_best_action(self, x: torch.Tensor) -> torch.Tensor:
return self._imitator(x).argmax(dim=1)