DLR-RM / rl-baselines3-zoo

A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.
https://rl-baselines3-zoo.readthedocs.io
MIT License
1.89k stars 494 forks source link

[Bug]: Optimization log and optimal policy not in `--optimization-log-path` but in `--log-folder` #444

Open turbotimon opened 2 months ago

turbotimon commented 2 months ago

🐛 Bug

The help for --optimization-log-path claims: "Path to save the evaluation log and optimal policy for each hyperparameter tried during optimization. "

However, when i run optimization with --log-folder logs/ and --optimization-log-path optimization/, the log (csv) and optimal policy (pkl) gets saved to logs/:

logs/
  a2c/
    report_MountainCar-v0_1-trials-1-tpe-median_1713858676.csv 
    report_MountainCar-v0_1-trials-1-tpe-median_1713858676.pkl
optimization/
  trial_0/
    best_model.zip
    evaluations.npz

So either the documentation/help for these parameters needs to be changed or the code where to write the report

To Reproduce

python train.py --algo a2c --env MountainCar-v0 -optimize  --log-folder logs/ --optimization-log-path optimization/

Relevant log output / Error message

No response

System Info

Checklist

turbotimon commented 2 months ago

(@araffin just fyi: If you have decided what you prefer (changing the help/doc strings OR fix where it will be saved), I'm happy to do a merge request)