muzero

An unofficial reimplementation of MuZero, Gumbel MuZero in Pytorch.

Requirements

The project is running on Python 3.10. To install dependencies, run the following command

pip install -r requirements.txt

Running experiments

Each experiment can be run by calling main.py, choosing mode (train or test), and either with required arguments or with a predefined config file with tag --config-path. For instance:

python main.py train --config-path configs/train/muzero_cartpole.json

Acknowledgements

The code is heavily inspired by these repos:

• muzero-general
• mctx

References

[1] Julian Schrittwieser, Ioannis Antonoglou, Thomas Hubert, David Silver et al. Mastering Atari, Go, chess and shogi by planning with a learned model. Nature 588, 604–609, 2020.
[2] Ivo Danihelka, Arthur Guez, Julian Schrittwieser, David Silver. Policy improvement by planning with Gumbel. ICLR, 2022