With tinyzero, you can easily train agents similar to AlphaZero on any environment you desire. To get started, make sure you have all the necessary dependencies installed by running “pip install requirements.txt”. To train an agent on an existing environment, use the command “python3 tictactoe2d/train.py”, replacing “tictactoe2d” with the environment name. You can modify parameters like the number of episodes and simulations in the train script. To evaluate the trained agent, run “python3 tictactoe2d/eval.py”. If you want to add a new environment, follow the game.py files in the existing examples and implement the required methods. Similarly, to add a new model or agent, refer to the examples provided in models.py and agents.py, respectively.
https://github.com/s-casci/tinyzero