Feat : Gigastep wrapper

instadeepai / Mava

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

Apache License 2.0

737 stars 90 forks source link

Feat : Gigastep wrapper #1027

Closed Louay-Ben-nessir closed 8 months ago

Louay-Ben-nessir commented 9 months ago

What?

Added a wrapper for the Gigastep environment.

Why?

to make the Gigastep environment Mava compatible.

How?

This wrapper takes in the output from the Gigastep environment and converts it into a Mava suitable input and vice-versa.

Extra

Gigastep is fully competitive. To make it cooperative the adversary team is taking random actions until a better solution is found. Only discrete actions are currently supported. Only vector observations are currently supported.

CLAassistant commented 9 months ago

All committers have signed the CLA.