Chess evaluation using machine learning model

Model created by Demid Efremov as part of PMLDL course

d.efremov@innopolis.university

The model aims to evaluate different chess positions using convolutional neural network

The project is using Lichess positions dataset, and comes together with a chess bot created on the same site. Check out bot's code in this repository.

Requirements

This projects uses pytorch as the machine learning network.

Requirements can be installed with:

$ pip install -r requirements.txt

Usage

1. Install the requirements
2. Download the model here, and unzip it into the project folder.
3. Run the inference script with $ python evaluate.py

Command line parameters can be seen with

$ python evaluate.py -h

Training

1. Install the requirements
2. Run the dataset downloader with $ python download_dataset.py
3. Run the training with $ python train.py

Use -h flag with any script to see configurable parameters. Check out the notebooks for extra info and code snippets.