s-casci/tinyzero
A framework for training AlphaZero-like reinforcement learning agents on custom environments via self-play and MCTS.
Not currently ranked — collecting fresh signals.
star history
The repository provides a streamlined implementation of the AlphaZero algorithm for training game-playing agents. It uses Monte Carlo Tree Search combined with deep neural networks for self-play training. Users can add custom environments by implementing a defined interface (reset, step, get_legal_actions, to_observation, etc.) and train agents through configurable episodes and simulations. It includes wandb integration for logging training metrics.
Frequently asked
- What is s-casci/tinyzero?
- A framework for training AlphaZero-like reinforcement learning agents on custom environments via self-play and MCTS.
- Is tinyzero open source?
- Yes — s-casci/tinyzero is open source, released under the MIT license.
- What language is tinyzero written in?
- s-casci/tinyzero is primarily written in Python.
- How popular is tinyzero?
- s-casci/tinyzero has 436 stars on GitHub.
- Where can I find tinyzero?
- s-casci/tinyzero is on GitHub at https://github.com/s-casci/tinyzero.