Updates to docs, environments, and bug fixes

Rohan138 commented 2 years ago

Changes:

Added conda_uint8_patch.sh
Added example scripts for training with Stable-Baselines3: run_scripts/sb3_train.py and run_scripts/sb3_independent.py
Refactored setup.py and updated requirements
Added the inequity_averse_rewards, alpha, beta arguments
Fixed pettingzoo wrapper and removed add_default_args from the environments themselves
Other misc fixes

Note: I'll be creating two more PRs in the near future (moved outside this PR for clarity) to:

Update to the latest RLlib version and (hopefully) make it so that we can remove both patch scripts and release the environments as a pip package
Switch MapEnv from RLlib's MultiAgentEnv to a PettingZoo ParallelEnv with an RLlib wrapper; this'll enable better integration with e.g. SB3, Tianshou and so on, while being fully backwards compatible

eugenevinitsky commented 2 years ago

Let me know when this is ready for review?

Rohan138 commented 2 years ago

This one's ready for review; all three PRs are self-contained, although updating the RLlib version is difficult because of how many changes there are.

eugenevinitsky commented 2 years ago

Yeah, that RLlib code is going to be incredibly hard to update given how much they have changed their API. Okay, I'll review this at some point during the holidays. Thanks for the PR!

jkterry1 commented 2 years ago

@eugenevinitsky it's been a month, so I'm just following up on this :)

eugenevinitsky commented 2 years ago

Approved! Thanks for this.

eugenevinitsky / sequential_social_dilemma_games

Updates to docs, environments, and bug fixes #202