DLR-RM / rl-baselines3-zoo

A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.
https://rl-baselines3-zoo.readthedocs.io
MIT License
2.01k stars 510 forks source link

Add documentation, switch to `pyproject.toml` and `ruff` #365

Closed araffin closed 1 year ago

araffin commented 1 year ago

Description

Add proper documentation because the README was a mess.

Also do the same as in https://github.com/DLR-RM/stable-baselines3/pull/1361

Can be see here: https://rl-baselines3-zoo.readthedocs.io/en/feat-doc/index.html

@qgallouedec in case you can do a quick review that's nice, otherwise I'll merge it without as this is important to provide clearer doc.

You don't need to go though all files, there are too many changes.

Motivation and Context

See #299

Types of changes

Checklist:

Note: we are using a maximum length of 127 characters per line

qgallouedec commented 1 year ago

Looks great! It's much better now! I didn't go through all the files since it's unlikely to break anything.

araffin commented 1 year ago

I didn't go through all the files since it's unlikely to break anything.

The build is somehow broken now, but related to setuptools I think.

qgallouedec commented 1 year ago

It seems like enjoy is broken for PandaReach. I'm taking a look

araffin commented 1 year ago

It seems like enjoy is broken for PandaReach. I'm taking a look

The problem comes from VecNormalize which overrides somehow the observation space. Will see if I can push a fix soon.

araffin commented 1 year ago

I found the issue, it is a mix of https://github.com/DLR-RM/rl-trained-agents/pull/10 and https://github.com/DLR-RM/stable-baselines3/commit/12e9917c24dc23d7de7694a924f017c6a8e9a6ce (where the observation space was set to the env one), I will push a fix for the VecNormalize stats.