Tunny v0.9.1
Summary
Preferential optimization
support artifact input
Added
- The seed value can be specified in BayesianOptimization(GP).
- Support optuna artifact feature.
- GeometryBase & FishPrint & file path are supported.
- Allows Trial to FAIL.
- In contrast to the constraint, if the value of the objective function cannot be obtained properly due to divergence of the solution or other reasons, it can be reflected in the trial.
- It works by inputting a bool value in the Attributes with the key "IsFAIL".
- Preferential optimization support.
- Human-in-the-loop optimization with PreferentialOptimization for single-objective images, and human-in-the-loop optimization with sliders for multi-objective images.
Changed
- Tunny Python runtime environment location to [UserProfile]/.tunny_env
- Moved settings file location to [UserProfile]/.tunny_env
- Storage's default path to the desktop.
- Previously, it was in the same folder as the Tunny component .gha file.
- Default StudyName set to empty.
- If you don't input a name in Tunny UI, study Name using GUID will be automatically inserted.
Fixed
- The seed value of the sampler was not used.
- Failure to launch Optuna-Dashboard from tab.
Related PR
- Fix/seed value is not used by @hrntsm in #198
- Update sample.gh by @reisyu in #204
- use tunny env folder by @hrntsm in #207
- Fix/tab optuna dashboard button by @hrntsm in #208
- Fix debug log path by @hrntsm in #209
- Add support artifacts by @hrntsm in #210
- Update requirements.txt & chagelog by @hrntsm in #211
- Feature/support trial fail by @hrntsm in #212
- Feature/preferential optimization by @hrntsm in #213
- Release v0.9.0 by @hrntsm in #214
New Contributors
Full Changelog: v0.8.2...v0.9.0