AlphaZero

A simplified AlphaZero implementation written in Python with TensorFlow and Keras.

The implementation is based on the article published to Nature by David Silver, Julian Schrittwieser, Karen Simonyan et al.

Here is a demonstration with Tic-Tac-Toe. AlphaZero with Tic-Tac-Toe

You can run the demonstration program on Kaggle (an account is required).

You can also download the file to run the program on your computer.

AlphaZero Tic-Tac-Toe Demo.zip

If you have docker, you can build and run the image by executing docker build and run.sh.

You may also run the program directly on your machine. To run the program, you should have the prerequisites listed below.

Softwares

  • NVIDIA CUDA 11.2 (Download)
  • Python 3.10

Python Packages

  • typing_extensions 4.12
  • NumPy 1.26
  • TensorFlow 2.10
  • Keras 2.10

Then, you can start with this command.

python -m games.mnk_game.executable.play