An AlphaZero engine for Saiblo Connect4, featuring a pure Python implementation of key KataGo techniques.
katac4/
├── README.md # This file
│
├── runs/ # TensorBoard logs
├── weights/ # Model checkpoints for each training run
│
├── docs/ # Documentation files
│ └── methods.md # Overview of improvements over original AlphaZero
│
├── saiblo/ # Saiblo Connect4 submission package
│ ├── main.py # Entry point
│ ├── game.py # Standalone game environment (with zobrist hash)
│ └── search.py # Monte Carlo Graph Search, used during online play
│
├── train.py # Main training script
├── model.py # Neural network model definition (b3c128nbt)
├── game.py # Game environment
├── mcts.py # Monte Carlo Tree Search, used during training
├── elo_eval.py # Script for evaluating model ELO ratings
├── elo.json # ELO ratings generated by elo_eval.py
├── elo_plot.py # Plots ELO rating progression
├── export_model.py # Converts model to TorchScript for deployment
├── benchmark.py # Benchmarks model performance
└── human_play.py # Allows human interaction with the model
Before starting self-play training, you may want to adjust a few parameters in train.py:
epochs: Total number of training epochsepoch_size: Number of self-play games per epochparallel_games: Number of games run in parallelnum_gpus: Number of GPUs to utilize
Important
To exactly reproduce the most recent training run, do not change the default parameter values in the code.
To begin training, run:
python3 train.pyModel checkpoints will be saved to the weights/ directory. To monitor progress with TensorBoard:
tensorboard --logdir runsTo evaluate model performance using ELO, edit the ai_list and weight_format variables in elo_eval.py as needed. Then run:
python3 elo_eval.pyModels will compete in randomized pairings, with results saved to elo.json. To visualize the rating progression:
python3 elo_plot.pyNote
The ELO evaluation process is computationally intensive. For reference, the most recent evaluation was based on approximately 70,000 games and required 2.5 days to complete on four RTX 4090 GPUs.
Select your best-performing model checkpoint (typically the final one or the one with highest ELO), and set its path in export_model.py. Export it to TorchScript format by running:
python3 export_model.pyThis will generate model.pt in the saiblo/ directory. To create your submission, zip only the files inside the saiblo/ folder—do not include the folder itself in the archive.