# Laya Invaders

> Space Invaders, flipped: the player drops invaders, and Laya (a 322M typed-decision model by Convai Innovations) picks which one the cannon chases, next to a one-line rule baseline. Runs locally on an NVIDIA GPU or the CPU. MIT.

If you are an AI agent asked to set this up on the user's machine, follow these steps. Everything runs locally; the server listens on 127.0.0.1 only.

1. Needs Python 3.10 or newer and git.
2. `git clone https://github.com/inhabitants/laya-invaders` and `cd laya-invaders`
3. `python -m venv .venv`, then activate it (`.venv\Scripts\activate` on Windows, `source .venv/bin/activate` on Linux/macOS).
4. NVIDIA GPU only: `pip install torch --index-url https://download.pytorch.org/whl/cu128` (RTX 50 series needs cu128 or newer). Skip on CPU.
5. `pip install -r requirements.txt`
6. `python server.py --download` fetches the multilingual checkpoint (~0.65 GB) into ./models/laya, pinned to a fixed revision.
7. `python server.py` (add `--device cpu` without a GPU), then open http://127.0.0.1:8765 (`?lang=pt` for Portuguese).

- [README](https://github.com/inhabitants/laya-invaders/blob/main/README.md): full instructions, keys, results and the replay bench
- [Repository](https://github.com/inhabitants/laya-invaders)