rail-berkeley / serl

SERL: A Software Suite for Sample-Efficient Robotic Reinforcement Learning
https://serl-robot.github.io/
MIT License
375 stars 42 forks source link

fix the naming for utd_ratio #59

Closed jianlanluo closed 5 months ago

jianlanluo commented 5 months ago

As we are using async update, the old naming utd_ratio doesn't apply anymore. To avoid confusion, we are renaming it to "critic_actor_ratio", the empirical value is set to 4, but can also change to anything that's suitable