Skip to content

vwxyzjn/envpool-cleanrl

Repository files navigation

EnvPool + CleanRL

This repo contains the source code of the CleanRL experiments presented in EnvPool: A Highly Parallel Reinforcement Learning Environment Execution Engine

If you like this repo, consider checking out CleanRL (https://github.com/vwxyzjn/cleanrl), the RL library that we used to build this repo.

Get started

Prerequisites:

Install dependencies:

poetry install

Atari

Train agents:

poetry run python ppo_atari.py

Train agents with experiment tracking:

poetry run python ppo_atari.py --track

MuJoCo

Train agents:

poetry run python ppo_continuous_action.py

Train agents with experiment tracking:

poetry run python ppo_continuous_action.py --track

Replicating results from the paper

See benchmark.sh

Citation

@article{envpool,
  title={EnvPool: A Highly Parallel Reinforcement Learning Environment Execution Engine},
  author={Weng, Jiayi and Lin, Min and Huang, Shengyi and Liu, Bo and Makoviichuk, Denys and Makoviychuk, Viktor and Liu, Zichen and Song, Yufan and Luo, Ting and Jiang, Yukun and Xu, Zhongwen and Yan, Shuicheng},
  journal={arXiv preprint arXiv:2206.10558},
  year={2022}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published