LIN556699/Sudoku-Dataset

---
license: apache-2.0
size_categories:
- 10M<n<100M
---

# Dataset Card: Sudoku Dataset

## Dataset Overview

- **Dataset Sources:**

1. [1 million Sudoku games](https://www.kaggle.com/datasets/bryanpark/sudoku)
2. [3 million Sudoku puzzles with ratings](https://www.kaggle.com/datasets/radcliffe/3-million-sudoku-puzzles-with-ratings)
3. [4 Million Sudoku Puzzles Easy-to-Hard](https://www.kaggle.com/datasets/informoney/4-million-sudoku-puzzles-easytohard)
4. [9 Million Sudoku Puzzles and Solutions](https://www.kaggle.com/datasets/rohanrao/sudoku)
5. ~2k miscellaneous scraped puzzles

- **Dataset Size:** 17M puzzles (16.7M for training, 300K for evaluation)

## Data Format

- **File Format:** Parquet files
- **Dataset Split:**
- `train`: 16.9M puzzles
- `valid`: 50K puzzles

## Dataset Attributes

The dataset contains the following attributes:

- `puzzle`: Represents the Sudoku puzzle configuration.
- `solution`: Corresponding solutions to the puzzles.
- `difficulty`: Indicates the difficulty level of the puzzles.
- `source`: Source identifier for the puzzle (e.g., dataset name or origin).

## Data Preprocessing

- **Handling Blanks:** Blanks in the puzzles are represented as '0'.

## Licensing and Legal Considerations

- The dataset is sourced from various Kaggle datasets, and each source may have its own licensing terms. Please refer to the respective Kaggle pages for licensing information.