Jiayi Weng

It should work if you change ActorProb/Critic to RecurrentActorProb/RecurrentCritic. But as many users reported (#486), it only "works" for the workflow/pipeline perspective instead of ML perspective unfortunately.

https://github.com/thu-ml/tianshou/issues/567#issuecomment-1065786212

You should return a state with type `dict[str, torch.Tensor]` where the tensor shape is `[bsz, len, ...]` (and it will be the same as input `state`). You are returning `dict[str,...

Are you interested in submitting a PR to support this?

could you repro the same issue inside docker container?

need to make sure `envpool/mujoco/dmc/utils.cc` has the correct permission. What does `ls -lh envpool/mujoco/dmc/utils.cc` give?

could you try `BAZELOPT=--verbose_failures make bazel-build`? See `Makefile` for actual command.

Should be relatively straight forward (less than 30 lines of code), would you like to submit a PR?

You need to change directory, try `cd /tmp` and `python -c "import envpool"`?

it's a python issue -- if you have a dir called `envpool` and performing `python -c "import envpool"`, it will import from that directory instead of `~/.local/.../python3.10/site-packages/envpool/`