Feat: support SMAX - Githubissues

instadeepai / Mava

🦁 A research-friendly codebase for fast experimentation of multi-agent reinforcement learning in JAX

Apache License 2.0

709 stars 83 forks source link

Feat: support SMAX #970

Closed OmaymaMahjoub closed 8 months ago

OmaymaMahjoub commented 9 months ago

What?

Support SMAX environment from jaxmarl

Why?

Add more environments that Mava supports

How?

Get action masking if exists
Get global state instead of concatenate obs for centralized training.
Get win rate during evaluation and log it to Neptune, json file, and cout in terminal
Extra

TODO: add 27m_vs_30m results to the readme