AZ-Go: A Custom AlphaGo Zero Implementation for Compact Board Configurations
A distributed AlphaZero implementation specifically designed for training Go models on a 7x7 board. This project implements the AlphaGo Zero algorithm with distributed training capabilities, neural network management, and comprehensive logging.
Authors: Toan Nguyen, Blake Good, Harrison Leath
Date: May 5, 2025
Project Resources
🔗 Notion Documentation & Model Archives
Access our Notion workspace containing:
- Model Archives - Download trained model checkpoints
- Training Logs - Detailed experiment records
- Performance Benchmarks - Model comparison data
- Research Notes - Implementation insights and findings
Documentation Structure
📁 Getting Started
- Installation Guide - Set up AZ-Go on your system
- Usage Guide - Run training and play games
🏗️ Technical Documentation
- Codebase Structure - Navigate the source code
- Architecture Overview - System design and components
🚀 Training & Deployment
- Training Process - Understanding the training pipeline
- Distributed Training - Scale across multiple machines
📊 Analysis & Research
- Model Analysis - Evaluate and understand your models
- Analysis Tools - Built-in analysis capabilities
- Related Papers - Research foundation and references
License
This project is licensed under the MIT License - see the LICENSE file for details.