TinyChess

Tiny chess engine for learning purposes using conv2d

Link to the Kaggle Notebook

https://www.kaggle.com/code/m2rads/tinychess?scriptVersionId=330189406

Weights: https://huggingface.co/m2rads/tinychess/tree/main

Setup

uv venv .venv
source .venv/bin/activate
uv pip install -r requirements.txt

cd frontend 
npm install
cd ..

fastapi dev

Then navigate to http://127.0.0.1:8000/

Model weights are stored on Hugging Face and downloaded automatically on first run.

Running training scripts interactively (REPL)

The training scripts (e.g. training/llm.py) support cell-by-cell execution using # %% cell markers, compatible with any Jupyter kernel-based REPL (VS Code, Zed, JupyterLab, etc.).

Setup:

uv pip install ipykernel huggingface_hub

Then in your editor, select the .venv Python environment as your kernel and run cells with the editor's REPL shortcut (shift+enter in most IDEs).