An confusion about sample

def sample(self, batch_size: int) -> Tuple[TEN, TEN, TEN, TEN, TEN, TEN]:
ids = th.randint(self.cur_size - 1, size=(batch_size,), requires_grad=False)
return (
self.states[ids],
self.actions[ids],
self.rewards[ids],
self.undones[ids],
self.unmasks[ids],
self.states[ids + 1],
)
Is this not a good way to deal with it? I think other people use dictionaries to deal with it。

Start by reading the sample method shown in the issue and tracing its callers to understand the expected six returned values and the state indexing. Compare that expectation with the project’s existing replay-buffer usage; done means the behavior or intended tuple structure is clarified and any necessary change is covered by the relevant tests, though no test file is named in the issue.

Rédigé par le modèle d'indexation à partir du texte de l'issue.