/gobang

Primary LanguagePython

Gobang

This is a PyTorch-based implementation of AlphaGo Zero algorithm for gobang game. Also, provide an OpenAI Gym style environment interface for gobang game

Environment

  • Python 3.*
  • PyTorch 0.3.*

Usage

Training

python train.py

Result

6*6 board, n_in_row is 4

  • training time

    2~3 hours

  • results

    9:1 versus pure monte carlo tree search algorithm(with 10 times sample counts)

TODO

  • Add interface for human player
  • Implement render_ interface for gobang environment so as to provide an UI