search

facebookresearch / rlmeta

RLMeta is a light-weight flexible framework for Distributed Reinforcement Learning Research.
MIT License
284 stars 28 forks source link

Refactor PPOAgent and PPORNDAgent #80

Closed xiaomengy closed 1 year ago

xiaomengy commented 1 year ago

This PR refactored PPOAgent and PPORNDAgent implementation.

  1. For PPOAgent, hide member variables as private.
  2. For PPORNDAgent, remove the "next_obs" field from replay_buffer to save memory and improve performance.