Datasets and Dataset Collections

Below is a collection of free / open datasets and lists of datasets at the interface between chemistry, materials and machine learning / AI.

Foundations

• Hugging Face

  Ecosystem of pretrained models, datasets, and Python libraries for transformers, diffusion models, and other modern ML architectures.

  License: -

  transformersLLMsmodel hub

Chemistry

Materials

• MatChem Dataset Repository

  Curated list of datasets for machine learning with materials, including links to data resources and related projects.

  License: MIT

  datasetsmaterialscurated list

• Digital Materials Foundry – Experimental Materials Data Library (Henry Royce Institute)

  Library of experimental materials data repositories curated by the Henry Royce Institute. Includes device-performance, stress–strain, thermoelectric, optical property databases, etc. Useful resource for ML in materials. License varies per dataset. oai_citation:0‡Henry Royce Institute

  License: Varies per dataset (MIT, CC-BY-4.0 etc.)

  experimental materials datamaterials discoverydata-library

• Materials Project

  Open-access computational materials database providing predicted and known properties of inorganic materials (e.g., formation energy, band-gap, structure) built via DFT and high-throughput workflows. Widely used in ML for materials. oai_citation:1‡Wikipedia

  License: CC-BY-4.0

  inorganic materials datasetDFT computed propertiesmaterials ML

• MatBench – Benchmark Datasets for Materials Property Prediction

  Benchmark dataset suite curated by the Materials Project for ML-based materials property prediction. Tasks range across electronic, thermal, mechanical properties; includes APIs & leaderboard. License: MIT. oai_citation:3‡GitHub

  License: MIT

  benchmark datasetsmaterials MLproperty prediction

• Porous Material AI Gym: Open Datasets for Machine Learning on Porous Materials

  Collection of open datasets for machine learning pertaining to porous materials (MOFs, COFs, zeolites). Includes thousands of labelled examples (adsorption, band-gaps, charges) for supervised learning. Provides ready-to-use data for ML workflows.

  License: -

  porous materialsmachine learning datasetMOFs/COFs