certen/Sudoku-1M-Puzzles

1M Sudoku Puzzles (9x9)

A dataset of 1,000,000 unique 9x9 Sudoku puzzles and their corresponding solutions. Each puzzle has between 40 and 60 blank cells, spanning various difficulty levels.

Dataset Structure

The dataset is provided in a single CSV file with the following columns:

• puzzle: 81-character string representing the initial board (0 = blank).
• solution: 81-character string representing the completed board.

Example Row

puzzle,solution
00000001040000000000000000... , 526837941413925768...

Data Generation

Puzzles were generated using a custom geometric transformation engine based on a seed board and then blanked out randomly. This method ensures all solutions are valid Sudoku boards.

Usage

This dataset is designed for training and benchmarking machine learning models (Transformers, CNNs, GNNs) on Sudoku solving tasks.

Loading with datasets library

from datasets import load_dataset

dataset = load_dataset("certen/Sudoku-1M-Puzzles")
print(dataset["train"][0])

License

MIT