metadata
configs:
- config_name: default
data_files:
- split: train
path: train/*.pgn.zst
- split: test
path: test/*.pgn.zst
license: mit
tags:
- code
size_categories:
- 100M<n<1B
pretty_name: ChessBot-Dataset
dataset_info:
features:
- name: state
dtype:
array2_d:
shape:
- 8
- 8
dtype: int8
- name: action
dtype: int16
- name: best_action
dtype: int16
- name: legal_actions
sequence: int16
- name: result
dtype: float32
splits:
- name: train
num_bytes: 65545518162
num_examples: -1
- name: test
num_bytes: 7298385021
num_examples: -1
download_size: 7621976329
dataset_size: 72843903183
This repository contains a gigantic curated chess dataset meant for machine learning. Currently the dataset contains approximately 2 billion positions in PGN format. Huge credits to the following main sources:
- Lumbra's Database (filtered 2600+)
- Lichess Puzzle Database
- Computer Chess: TCEC Database, CCRL, LC0 training data (test80)
The supporting dataset code and training platform, known as ChessBot-Battleground, is found at: https://github.com/KeithG33/ChessBot-Battleground/
Note: The dataset uses a custom loading script (ChessBot-Dataset.py) to parse PGN data on the fly.