Skip to content

kevinkorrol/DeepFork

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DeepFork

Repo authors: Tõnis Lees, Kevin Korrol

Project motivation and goal

The goal of this project was to learn about chess engines and create and train a model that would make better moves than random. We decided that the project would be a success if our chess engine can beat the lowest rated bot on chess.com website, called Martin. In case the model works really well, it could be strong enough to beat an average player easily.

Repo Guide

/

get_data scripts are used to get all the games from the pgnmentor.com page and combine them into one processed .pgn file.

/data

Folder for storing .pgn and .pt files:

  • /processed - .pt files to store input tensors
  • /raw - .pgn files for raw unprocessed data

/models

Folder for storing trained models.

/notebooks

.ipynb files that are not part of the main project flow but were used to analyze the the training data.

/src

General folder containing files used to play chess against a trained model, train models, etc.

/utils

Contains helper functions for the whole workflow.

Replicating results

The first step would be to fetch the .png data from the internet using the get_data script. After that, the .pgn files need to be converted into .pt files that can be used as input for the model. When that is done, run train.py to train the model.

HW10: https://docs.google.com/document/d/12dOD-GB6gpn7qroGx5NYyiwk3bp99Q19QCCgOZOlU38/edit?usp=sharing

About

Neural net that sees every fork coming

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •