Genesis-1

A local web UI for playing against the Genesis-1 chess engine, an AlphaZero-style model trained from self-play. It's a 15-block / 192-channel residual network (~10M parameters).

Quick start

1. Download the .pt file from HuggingFace
2. Place it in the models/ folder.
3. Install dependencies:

   pip install -r requirements.txt

4. Start the server:

   python app.py

5. Open http://127.0.0.1:8000 in your browser.

Flags

• --model: Load a model manually (default: models/genesis_1.pt)
• --device: Choose between the cpu or the gpu (options: cpu|cuda; default: auto detect)
• --host 0.0.0.0: Expose on LAN (default is loopback only)
• --port: Change the port (default 8000)
• --seed: RNG seed for the engine's move selection (default: 0). Set it for reproducible games.

Exposing the server beyond loopback: The default --host 127.0.0.1 is safe because only your machine can reach it. If you bind to 0.0.0.0 (or another network interface), anyone who can reach the port can submit arbitrary inputs to the model. There's no auth and no rate limit.

Strength levels

The strength slider in the UI maps to MCTS simulations per move:

Level	Sims
Weak	100
Medium	200
Strong	400
Strongest	800

More sims = stronger play, longer think time.

Training

Genesis-1 was trained from self-play on:

• Hardware: GPU: RTX 3090 | CPU: Ryzen 9 5950X | RAM: 32GB
• OS: Windows 11
• Training time: ~12 days