Skip to content

rvbansal/muzero_tf

Repository files navigation

MuZero

A TensorFlow implementation of DeepMind's MuZero algorithm for self-learning games without any knowledge of the rules. The algorithm is implemented as described in the original paper and pseudocode. It supports prioritized replay and is parallelized with the help of Ray. The repo structure is based on a muzero-pytorch.

Train: python main.py --mode train --env CartPole-v1 --force

Test: python main.py --mode test --env CartPole-v1 --force

TensorBoard: tensorboard --logdir=result_dir

At the moment, the code has only been tested for simple OpenAI gym environments like CartPole. Results are fairly sensitive to choices of hyperparameters.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages