Antonin RAFFIN

> I'm glad to stash this conversation (and work done), it still might be valuable for future design conversations. I think I was not so clear, please open a draft...

I think @kachayev did not have time to do the PR. > action_logits = th.mul(action_logits ,self,mask_matrix) you should probably take a look at `MaskablePPO` in our contrib repo: https://sb3-contrib.readthedocs.io/en/master/modules/ppo_mask.html

> I am wanting to make a wrapper around PySC2 now that dictionary observations are supported, however multiple output policy support is still required. what type of multi output policy...

Hello, >stable_baselines3.common.preprocessing.get_obs_shape returns the wrong shape when a MultiBinary spaces is multi-dimensions. quick question: why would do `spaces.MultiBinary([5, 4, 5])` instead of `spaces.MultiBinary(14)`? (as each dimension is independent) (I know...

it seems that the failures come from `pygame` not being installed somehow, probably an issue from gym...

> won't be available for SB3 until the next release by Gym? https://github.com/openai/gym/issues/2640 must be fixed first too

> Before gym 0.22.0, the default seed() method was not actually setting any seed [[source](https://github.com/openai/gym/blob/v0.21.0/gym/core.py#L124)]. In gym 0.22 the default seed() does set the seed [[source](https://github.com/openai/gym/blob/2dddaf722acccfd0412d745890c40dcd972586d5/gym/core.py#L157)]. I see... > is...

> but now every time env.seed is called will raise a warning. Could we make the `VecEnv` seed method backward compatible? In the sense that if `seed` is accepted in...

@carlosluis i think there was a misunderstanding with my remark here: https://github.com/DLR-RM/stable-baselines3/pull/780#issuecomment-1062235112 please do not change `VecEnv` signature but make `VecEnv` `seed()` method forward compatible (so it supports all gym...

> AFAIK, to make VecEnv seed() forward compatible we should internally call reset(seed=seed) for every env, similar to how it's now done in DummyVecEnv. yes > but I'm not sure...