Extension for Scikit-learn* is happy to introduce 2026.0.0 release!
🚩 Removals and ABI Compatibility
- The following functionality is removed in the 2026.0.0 release:
- Fitted attributes of BasicStatistics without underscores
- The ‘use_raw_inputs’ option from GPU estimators scikit-learn-intelex. Array API should be used instead
- Support for deprecated DPCTL tensors. DPNP arrays should be used instead
- The deprecated option to build scikit-learn-intelex without the onedal interface, increasing the minimum version of oneDAL required
🚨 What's New
- Introduced new Extension for Scikit-learn* functionality:
- t-SNE can now work with sparse data
- Sparse pandas data frames are now correctly handled as sparse inputs
- Patching for individual estimators now uses full paths and names of the patched scikit-learn objects to reduce ambiguity
- Added preview support for LogisticRegressionCV
- Array API support is expanded to more estimators (KNN, K-Means, LogisticRegression)
🪲 Bug Fixes
- Fixed thread-safety issues in GPU estimators
Acknowledgements
Thanks to everyone who helped us make 2026.0.0 release possible!
@Alexandr-Solovev, @Vika-F, @david-cortes-intel, @icfaust, @napetrov, @maria-Petrova, @homksei, @ahuber21, @ethanglaser, @razdoburdin, @avolkov-intel, @yuejiaointel, @DDJHB, @kjackiew, @richardnorth3
Full Changelog: 2025.11.0...2026.0.0