AlphaZero

Machine learning project to recreate, in smaller scale, the AlphaZero algorithm by Google Deepmind. Testing the algorithm on Tic-Tac-Toe, Connect4, and Checkers.

Setup

To run the program correctly you need to install:

pip install autopep8
pip install progressbar2

You need to have a Nvidia gpu, you can install the gpu version of tensorflow

pip install tensorflow-gpu

and all the related requirements, such as Cuda and CuDnn.

Before running the program, create an empty folder named 'saved' if not already there.

Running

There are different modes that can be used. For training mode do:

python3 training.py
python3 main.py

for all other moves, you need to change the mode in the code at line 193 to either 'evaluation', 'debug' or 'manual'