Dependency Update Release

Maintenance release that supports newer versions of dependencies.

• Add support for Python 3.12
• Add support for Numpy>=2.0
• Support newer versions of dependencies:
  • imbalanced-learn: 0.13
  • XGBoost: 3.0
  • lightgbm: 4.6
  • SnapML: 1.16
  • scikit-learn: 1.6
• Improvements to test datasets
  • Replace use of no longer available dataset
  • Enable support for prefetching and caching datasets in CI
• Update all static tools

Removals/Known Issues:

• Remove support for Python 3.8 and Python 3.9
  • Python 3.9 is still believed to work , but it is no longer tested as part of CI and hence no longer supported.
• Disable testing for RASL subsystem.
  • It should still work with older versions of Numpy and Python, however it currently breaks with newer versions
  • This breakage will hopefully be fixed, and tests re-enabled, in a future release.
• Testing now generally happens against Numpy>=2.0. Numpy 1.X should still work, however it is no longer well-tested or supported.

Version 0.8.4

Adds support for hyperopt 0.2.7

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