Fix win rate calculation and update smax logging

FLAIROx / JaxMARL

Multi-Agent Reinforcement Learning with JAX

Apache License 2.0

425 stars 79 forks source link

Fix win rate calculation and update smax logging #28

Closed benellis3 closed 1 year ago

benellis3 commented 1 year ago

The win rate was calculating the global average win rate through training before. This update only logs and updates the win rate for the most recent episode.