Croissant: a metadata format for ML-ready datasets
Not a pilot. Kaggle, Hugging Face, and OpenML just adopted Croissant—a new standard for ML datasets that TensorFlow, PyTorch, and JAX can now load natively.
Why it matters
Google and MLCommons launched Croissant, a metadata format addressing a critical friction point in ML development: the fragmentation of dataset formats. With immediate adoption across major platforms and frameworks, this standardization could meaningfully accelerate ML model training and reduce time spent on data preparation.
The key facts
6 to knowCroissant format adopted by Kaggle, Hugging Face, and OpenML
Compatible with TensorFlow, PyTorch, and JAX via TensorFlow Datasets package
Built on schema.org, already used by 40M+ datasets
Includes Responsible AI (RAI) vocabulary extension for compliance, fairness, and explainability
Community collaboration includes Meta, NASA, Harvard, and 15+ other institutions
Includes open-source Python library, visual editor, and Dataset Search integration
Go to the source
Google Research Blogblog.research.google
Publisher excerpt: Posted by Omar Benjelloun, Software Engineer, Google Research, and Peter Mattson, Software Engineer, Google Core ML and President, MLCommons Association Machine learning (ML) practitioners looking to reuse existing datasets to train an ML model often spend a lot of time understanding the data,…