Releases: IBM/lale
Releases · IBM/lale
Version 0.8.4
Version 0.8.3
- Visualizations as SVG with links and tooltips in jupyter notebook >=7.
- Eliminated dependency on numpy<2 directly in Lale, but note that some packages that are used with Lale still depend on numpy<2.
Version 0.8.2
- Improvements to the release deployment scripts
Version 0.8.1
This release includes minor bug fixes
- Fix for xgb classifier without a label encoder
- URL links fix to open them in new tab
- Use newer AIF360 version that does not require an upper bound on numpy version
Version 0.8.0
- Add support for Python 3.11
- The SMAC hyperoptimization backend is not currently supported on 3.11
- Add support for scikit-learn 1.3 and 1.4
- Update supported versions of numpy, scipy, pandas, xgboost, lightgbm, snapml, and tensorflow
- AIF360 does not currently support numpy>=1.24, so testing for aif360 libraries is done using an older version of numpy
- Fix readthedocs build so our documentation is correctly updated
- Add label encoding to our wrapper for XGBClassifier, since it longer does this.
- Shuffle data in some batching tests to (mostly) avoid problems with encoding
- Remove support for Python 3.7 and scikit-learn < 1.0
- Remove autoai_libs wrappers and tests (they have been moved into the autoai_libs package).
Version 0.7.11
Fix a bug where links from pipeline visualizations to operator documentation did not work.
Version 0.7.10
- guard import of mystic
- add upper bound to tensorflow dependency
- remove protobuf installation version constraint.
Version 0.7.9
- Relax some dependency upper bounds
- Add RandomUnderSampler from imblearn
- Add Orbis and Urbis Mystic implementations
- Add 7 more fairness datasets
- Documentation improvements
- Add wrapper for autoai_libs.nsfa transformer
Version 0.7.8
- Updates to autoai_libs schemas to improve pretty-printing
- Update version of static checkers
Version 0.7.7
- Improvements to CI
- more static checking
- only using trusted actions
- Improvements to aif360 schemas and derived operators
- Update SnapML schemas for new version
- loosen scipy release bounds