Denys88 / rl_games

RL implementations
MIT License
941 stars 155 forks source link

Small changes for DexPBT paper. #204

Closed alex-petrenko closed 2 years ago

alex-petrenko commented 2 years ago

1) load env state from checkpoint in player (i.e. if we train with curriculum we want to load the same state). 2) Minor speedup disabling validate_args in distributions 3) added another way to stop training max_env_steps