chessRL

Setup

1. Setup Filestructure

Install Dependencies

pip install -r requirements.txt

Create data and weights folders

Create a data folder from the root of this project. Create two folders, one for logs and one for weights inside rl_chess.

1. Download Stockfish

Download the appropriate Stockfish binaries inside the data folder:

Linux:

wget https://github.com/official-stockfish/Stockfish/releases/latest/download/stockfish-ubuntu-x86-64-avx2.tar

tar -xvf stockfish-ubuntu-x86-64-avx2.tar

MacOS:

wget https://github.com/official-stockfish/Stockfish/releases/latest/download/stockfish-macos-m1-apple-silicon.tar

2. Download Puzzles

Download the puzzles from https://database.lichess.org/lichess_db_puzzle.csv.zst

From the data folder:

wget https://database.lichess.org/lichess_db_puzzle.csv.zst

3. Build Puzzles

Specify the path to the puzzle file in build_puzzles.py and then run python build_puzzles.py

Train From Self-Play

To train an RL agent to play by playing against itself, run python main.py self

Train With Puzzles

To train an RL agent to play from puzzle scenarios, run python main.py puzzle

Train Against Stockfish

To train an RL agent to play from full games against the Stockfish engine, run python main.py engine

Models

1. Self-Play
2. Engine
3. Puzzle
4. Self-Play + Puzzle
5. Engine + Puzzle
6. Puzzle + Self-Play
7. Puzzle + Engine