GNU Go¶
GNU Go is a classical Go engine that does not use neural networks. It is a lighter alternative to KataGo, useful when GPU resources are limited or when a weaker opponent is desired for curriculum training.
Installation¶
# Ubuntu / Debian / WSL
sudo apt-get install -y gnugo
# macOS
brew install gnugo
Service Module¶
The GNU Go integration lives in:
gym_gui/services/go_ai/gnugo_service.pygym_gui/services/go_ai/gtp_engine.py(shared GTP base class)
GNU Go communicates via the GTP (Go Text Protocol), sharing the same base class as KataGo.
Usage in MOSAIC¶
Human vs GNU Go: Play Go against a classical engine using the GUI.
Curriculum training: Use GNU Go as an easier opponent before graduating agents to play against KataGo.