Hugging Face Skops: new library to host scikit-learn models on HF Hub with model cards and inference widget

AI Impact Summary

Hugging Face is unveiling Skops to publish scikit-learn models on the HF Hub, including environment metadata and model cards to improve reproducibility and discoverability. The workflow leverages hub_utils and Card for packaging, metadata extraction, and pushing to the Hub, with inference widget support to enable runtime predictions. This tight integration with scikit-learn and the HF Hub lowers the barrier for production deployment of tabular-classification models and could shift teams toward HF Hub as a centralized model registry. Teams should plan for adopting Skops in pipelines that serialize models with pickle or joblib and ensure compatibility with the generated model cards and metadata.