AlphaZero Gomoku

A multi-threaded implementation of AlphaZero

Features

• Easy Free-style Gomoku
• Multi-threading Tree/Root Parallelization with Virtual Loss and LibTorch
• Gomoku, MCTS and Network Infer are written in C++
• SWIG for Python C++ extension
• Update 2019.7.10: Supporting Ubuntu and Windows
• Update 2022.4.4: Re-compile with CUDA 11.6/PyTorch 1.10/LibTorch 1.10(Pre-cxx11 ABI)/SWIG 4.0.2

Args

Edit config.py

Packages

• Python 3.6+
• PyGame 1.9+
• CUDA 10+
• PyTorch 1.1+
• LibTorch 1.1+ (Pre-cxx11 ABI)
• SWIG 3.0.12+
• CMake 3.8+
• MSVC14.0+ / GCC6.0+

Run

# Compile Python extension
mkdir build
cd build
cmake .. -DCMAKE_PREFIX_PATH=path/to/libtorch -DPYTHON_EXECUTABLE=path/to/python -DCMAKE_BUILD_TYPE=Release
make -j10

# Run
cd ../test
python learner_test.py train # train model
python learner_test.py play  # play with human

Pre-trained models

Trained 2 days on GTX TITAN X (similar to GTX1070)

See GitHub Release: https://github.com/hijkzzz/alpha-zero-gomoku/releases

GUI

References

1. Mastering the Game of Go without Human Knowledge
2. Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm
3. Parallel Monte-Carlo Tree Search
4. An Analysis of Virtual Loss in Parallel MCTS
5. A Lock-free Multithreaded Monte-Carlo Tree Search Algorithm
6. github.com/suragnair/alpha-zero-general