This is a PyTorch-based implementation of AlphaGo Zero algorithm for gobang game. Also, provide an OpenAI Gym style environment interface for gobang game
- Python 3.*
- PyTorch 0.3.*
Training
python train.py
-
training time
2~3 hours
-
results
9:1 versus pure monte carlo tree search algorithm(with 10 times sample counts)
- Add interface for human player
- Implement render_ interface for gobang environment so as to provide an UI