Antonin RAFFIN

Hello, This seems to be an issue very specific to a given problem. I would advise you to derive a `CustomDQN` from `DQN` to fit your needs (you could also...

>the callback doesn't have access to for DQN, it does through `self.locals`, for others, you need to wrap it.

Hello, It sounds like a callback should be the right solution (once #787 is merged) as you have access to `self.model`, you can call `self.model.replay_buffer.add()` inside the callback. In fact,...

Hello, >Is it planned to add Multidiscrete obs a spaces since DQN support them? This is not planned as we are focusing on the v3.0 for now (and avoid adding...

Hello, Did you try taking at look at the doc on [exporting models](https://stable-baselines.readthedocs.io/en/master/guide/export.html)? Btw, if you succeed, we would appreciate a PR that documents how to do it ;)

Looking at the policy (in `common/` folder), this should be: `'action': model.act_model._policy_proba` (cf https://github.com/hill-a/stable-baselines/issues/474) which corresponds to the output of the policy. `action_ph` is used for training (it is a...

Hello, In fact I encountered the same issue with Atari games... >Can I make the proposal to add a max_timesteps safety valve? Sounds like a simple and acceptable solution ;)...

>Is this expected behavior, and/or should LSTMs be tested in a different environment? I don't have much time to invest into that issue but if I get it right, there...

> Question was on how to test the current LSTM implementation if it works right, and so far there was trouble to solve a simple recall environment. @Miffyli We have...

The only maintainer that worked a bit with LSTM is @erniejunior , see https://github.com/hill-a/stable-baselines/issues/278 and https://github.com/hill-a/stable-baselines/issues/158