optuna 0.12.0

v0.12.0

on Python PyPI

This is the release note of v0.12.0. See here for the complete list of solved issues and merged PRs.

New Features

• Implement PercentilePruner. (#397, thanks @c-bata!)

Bug Fixes

• Fix the error that occurs when using a study that contains old trials. (#415)
• Synchronize when _ChainerMNTrial sets values. (#413)
• Fix a bug that FixedTrial cannot handle categorical parameters correctly. (#402)

Compatibility

• Make the step argument of should_prune method optional. (#398, thanks @c-bata!)

Documents

• Fix typo of LightGBM. (#409, thanks @nmasahiro!)
• Use optuna.load_study() function instead of optuna.Study() constructor. (#407)

Examples

• Fix FutureWarnings of scikit-learn. (#418, thanks @c-bata!)

Others

• Update flake8 exclude option in setup.cfg. (#405, thanks @c-bata!)