Feat: Support Continuous environments with unvmapped Nets

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

Closed WiemKhlifi closed 7 months ago

WiemKhlifi commented 7 months ago

Full support for continuous actions in all Mava PPO systems.

Add support for environments with continuous actions.

ContinuousActionHead added
Support for TensorFlow probability instead of distrax.
Access num_actionsand num_agents directly from environments (whether continuous or discrete).

This is the non-vamped version of the PPO networks.