rl-baselines3-zoo

Why we included this project

Anyone doing serious work with Stable Baselines3 will probably end up using this project. It bundles the practical parts of the RL workflow: training scripts, evaluation, video recording, and plotting, plus tuned hyperparameters for common environments and algorithms. The hyperparameters live in YAML files you can copy and adjust as a starting point, and the included pre-trained agents let you watch a well-configured policy behave before spending compute on your own training runs. That makes it handy for benchmarking algorithms and reproducing results, since you get sane defaults and reproducible configs in one place. It is also a decent way to learn how hyperparameter choices translate into RL performance.