werner-duvaud/muzero-general

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2861 stars · 671 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

A commented and documented PyTorch implementation of DeepMind's MuZero reinforcement learning algorithm, designed to be easily adapted to any game or Gym environment by adding a game file with hyperparameters. It is primarily intended for educational purposes and supports single and two-player games, multi-GPU training, and TensorBoard monitoring.

Use cases

• train a MuZero agent on chess or Go
• apply model-based reinforcement learning to a custom Gym environment
• learn how MuZero works from a documented implementation
• run self-play training on Atari games
• experiment with Monte Carlo tree search and learned environment models
• add a new game to a reinforcement learning framework

When to choose

• you want a readable, educational MuZero implementation to study or extend
• you need to adapt a state-of-the-art RL algorithm to a new game or Gym environment
• you want multi-GPU or Ray-cluster support for self-play training
• you prefer PyTorch and TensorBoard monitoring

When to avoid

• you need a production-grade, highly optimized RL library for large-scale training
• you want a simple plug-and-play solution without writing a game configuration file
• you need algorithms other than MuZero/AlphaZero-style model-based RL

Facets

library · maturity active

reinforcement-learning machine-learning deep-learning simulation reinforcement-learning machine-learning deep-learning python windows muzero alphazero mcts monte-carlo-tree-search model-based-rl pytorch gym self-play educational game-development linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="werner-duvaud/muzero-general")