A big thanks to Alexander Serkov for the web implementation.
models/TORCH_100EPOCHS has shown a performance of approx. 1500 ELO during opening and middlegame.
It failes after about 20 moves. Simple algorithm to detect blunders is needed.
-
Install Python dependencies:
pip install -r requirements.txt
-
Put your data (.pgn files) into
data/pgn/
.
The dataset that I used.
-
Refer to
engines/chess_engine.ipynb
for further instructions and actions (TensorFlow) -
Or see
engines/torch/predict