Michael Panchenko

Related to #949

The whole MARL stuff is a bit deprioritized now, but if you feel like it's an easy fix in the documentation, I'd be happy to review a PR :)

The problem is the output of the Actor - in your case it will be passed to torch's `Categorical` distribution. Something is going wrong in the forward. I can't spot...

We can't address all issues related to this, but the batch of improvements that can be done in version 1.x.y will be done in #1032

@Trinkle23897 we plan to address it after the high-level interfaces from @opcode81 are merged. If you have any other proposals, would be happy to hear them! Existing hpo approaches include:...

We will do this in (at least) two stages. The first will be a proper test-evaluation protocol for a single params config. @bordeauxred is on it

Thank you for the detailed description, I'm gonna think about this tomorrow and give a proper answer. But already now the situation you describe seems unusual - during training you...

We'll have to do #1058 first. The Collector is too convoluted to be touched, so we'll improve the situation in multiple steps

Right, would probably be useful. Gonna look into it soon

@opcode81 does this still make sense? You have recently introduced extensions to the atari/mujoco helpers