search

instadeepai / Mava

🦁 A research-friendly codebase for fast experimentation of multi-agent reinforcement learning in JAX
Apache License 2.0
700 stars 83 forks source link

Use rlax #929

Closed sash-a closed 6 months ago

sash-a commented 1 year ago

Feature

Use rlax where we can, so that we don't have to implement and test common methods.

Proposal

As far as I can see we can use it for policy/critic loss and GAE.

Benchmarking

sash-a commented 6 months ago

Investigated, it massively slowed down existing systems :disappointed: