search

Toni-SM / skrl

Modular reinforcement learning library (on PyTorch and JAX) with support for NVIDIA Isaac Gym, Omniverse Isaac Gym and Isaac Lab
https://skrl.readthedocs.io/
MIT License
443 stars 43 forks source link

Do not save checkpoints for eval runs #154

Open lopatovsky opened 3 weeks ago

lopatovsky commented 3 weeks ago

Motivation: Currently both train and eval modes are saving checkpoints, and best checkpoint. This was causing some troubles on our side when we implemented parallel validation runs that were then overwriting the checkpoints from the train run.

This change is passing a boolean flag save_checkpoint into post interaction function, which is set to False in case it is called from eval mode.