Skip to content

Home

Welcome to CleanQRL

docs Code style: black Imports: isort

CleanQRL is a Reinforcement Learning library specifically tailored to the subbranch of Quantum Reinforcement Learning and is greatly inspired by the amazing work of CleanRL. Just like the classical analogue, we aim to provide high-quality single-file implementations with research-friendly features. The implementations follow the ideas of CleanRL. They are clean and simple, yet scale nicely through additional features such as ray tune. The main features of this repository are

  • 📜 Single-file implementations of classical and quantum versions of 4+ Reinforcement Learning agents
  • 💾 Tuned and Benchmarked agents (with available configs)
  • 🎮 Integration of gymnasium, mujoco and jumanji
  • 📘 Examples on how to enhance the standard QRL agents on a variety of games
  • 📈 Tensorboard Logging
  • 🌱 Local Reproducibility via Seeding
  • 🧫 Experiment Management with Weights and Biases
  • 📊 Easy and straight forward hyperparameter tuning with ray tune

What we are missing compared to CleanRL:

  • 💸 Cloud Integration with docker and AWS
  • 📹 Videos of Gameplay Capturing

You can read more about CleanQRL in our upcoming paper.

Contact and Community

We want to grow as a community, so feel free to post Github Issues and PRs! If you are missing any algorithms or have a specific problem to which you want to tailor your QRL algorithms but fail to do so, you can also create a feature request!

Citing CleanQRL

If you use CleanQRL in your work, please cite our [paper]:

Coming Soon!

Citing CleanRL

If you used mainly the classical parts of our code in your work, please cite the original CleanRL paper:

@article{huang2022cleanrl,
  author  = {Shengyi Huang and Rousslan Fernand Julien Dossa and Chang Ye and Jeff Braga and Dipam Chakraborty and Kinal Mehta and João G.M. Araújo},
  title   = {CleanRL: High-quality Single-file Implementations of Deep Reinforcement Learning Algorithms},
  journal = {Journal of Machine Learning Research},
  year    = {2022},
  volume  = {23},
  number  = {274},
  pages   = {1--18},
  url     = {http://jmlr.org/papers/v23/21-1342.html}
}