Pgx
A collection of GPU-accelerated parallel game simulators for reinforcement learning, built with JAX.
What is Pgx?
Pgx is a library of hardware-accelerated game simulators for reinforcement learning, built natively in JAX. It provides fast, parallelized environments for classic board games and other discrete games, enabling efficient training of RL agents at scale. The project addresses the need for high-performance simulators in discrete action spaces, complementing continuous-space engines like Brax.
Reinforcement learning researchers and practitioners who need fast, parallel environments for training agents on board games or discrete games. It's particularly useful for those using JAX-based ML stacks and seeking GPU/TPU acceleration.
Pgx offers significantly faster simulation speeds compared to CPU-based alternatives by leveraging JAX's parallelization capabilities. Its JAX-native design ensures seamless integration with JAX-based RL algorithms and automatic differentiation, while maintaining compatibility with popular RL APIs like PettingZoo.
Overview
♟️ Vectorized RL game environments in JAX
Use Cases
Best For
- Training reinforcement learning agents on classic board games like Chess and Go
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
