google-deepmind/mctx
A JAX-native library implementing Monte Carlo tree search algorithms including AlphaZero and MuZero for reinforcement learning.
Not currently ranked — collecting fresh signals.
star history
Mctx provides JAX-native implementations of Monte Carlo tree search algorithms such as AlphaZero, MuZero, and Gumbel MuZero. The library enables efficient batched search operations with full JIT-compilation support, leveraging hardware accelerators for high-throughput inference. It is designed to work with deep neural network parameterized environment models, making it suitable for advanced reinforcement learning research and applications.
Frequently asked
- What is google-deepmind/mctx?
- A JAX-native library implementing Monte Carlo tree search algorithms including AlphaZero and MuZero for reinforcement learning.
- Is mctx open source?
- Yes — google-deepmind/mctx is open source, released under the Apache-2.0 license.
- What language is mctx written in?
- google-deepmind/mctx is primarily written in Python.
- How popular is mctx?
- google-deepmind/mctx has 2.6k stars on GitHub.
- Where can I find mctx?
- google-deepmind/mctx is on GitHub at https://github.com/google-deepmind/mctx.