Clarification Needed on num_depot Usage in MDCPDP Environment

[Moonbohoon]

I'm currently working with the MDCPDPEnv environment from this repository and encountered an issue regarding the usage of num_depot. In the _step and _get_reward function of env.py and render.py, num_depot is derived from td["capacity"].shape[-1], which seems unusual to me. Here's the relevant code snippet:

num_depot = td["capacity"].shape[-1]

I believe it should be derived from depot, not capacity. Is this understanding correct?
Any guidance or confirmation on this issue would be greatly appreciated. Thank you!
Best regards,

[Moonbohoon]

any update in here?

[699Felix]

I modified the code in class MDCPDPGenerator:
capacity = torch.randint( self.min_capacity, self.max_capacity + 1, # size=(*batch_size, 1), size=(*batch_size, self.num_depot), )
then you can see more depots not only one
but since here i cannot start the trainning of the MVMoE_POMO model……