Overview

This Gomoku AI is based on a Korea ML engineer's work on Github, but I'm can't find the original link when I finished it.

The difference to the previous work is:

• use pytorch rather than tensorflow
• use less training data, cause I don't have much RAM
• use pygame as GAME GUI rather than HTML

Install

Install numpy, pytorch, pygame library, then run:

python main.py

press r to restart game, q to quit game.

Training

Training process can be find on below kaggle notebook, training output is model.pth.

https://www.kaggle.com/code/metseq/gomoku-ai

Limits

I have not write the win condition, so the game will go on if you do not restart or quit.