Back to Gap Map Infrastructure / Classic ML & computer vision

skrub

skrub
open source / Overall score: 2.6

skrub is a Python library that prepares dataframes for machine learning with scikit-learn. Its TableVectorizer turns mixed numeric, categorical, date and text columns into features, with encoders such as GapEncoder, MinHashEncoder and StringEncoder for high-cardinality strings, and tabular_pipeline wraps that into a ready baseline. TableReport explores a table interactively. Development is supported by Inria and :probabl.

Openness

5 high confidence
5.0
license
BSD-3-Clause (OSI)
source
public (github.com/skrub-data/skrub)
core features withheld
no — community project with financial support from Inria and :probabl.; nothing sold

BSD-3-Clause and built from the public repository. skrub is a community project funded by Inria and the company :probabl., and nothing is sold beside it.

Adoption

3 high confidence
3.0

Measured on monthly PyPI downloads of the skrub package.

Capability

2 medium confidence
2.0

skrub turns raw tables into features and assembles a baseline pipeline around a scikit-learn estimator, but the learners themselves come from scikit-learn. It is a convenience layer over part of that toolkit rather than a catalog of models of its own.

  • https://skrub-data.org/stable/ recorded 2026-11-12

    skrub site: "skrub is a Python library to ease preprocessing and feature engineering for tabular machine learning"; "Create strong scikit-learn pipeline baselines effortlessly with TableVectorizer and tabular_pipeline"; "Encode text and high cardinality categorical data (StringEncoder, LLMEncoder, GapEncoder, and MinHashEncoder)".

Verified 2026-11-12