TUDataset: A collection of benchmark datasets for learning with graphs

Source: Dev.to

Overview

A large, free library of graph data designed to simplify and accelerate the development and testing of models that learn from networks.

• 120+ datasets covering a wide range of real‑world problems, suitable for both small‑scale and larger experiments.
• Ready‑to‑use example code, easy data loaders, and baseline implementations to avoid “setup hell.”
• Open source site and codebase, enabling reproducible results and straightforward sharing.
• Designed for fair comparison, faster progress, and broader participation across researchers, students, and hobbyists.

The collection aims to provide easy tools and open code for honest, reproducible experimentation, making it ideal for teaching, testing, or exploring new ideas in graph learning.

Read the comprehensive review:
TUDataset: A collection of benchmark datasets for learning with graphs