What's Changed * add warm start attribute to HpBandSterSearchCV, so that the user can choose whether to warm start by poetair in https://github.com/Yard1/hpbandster-sklearn/pull/7 * Use the best resource value instead of the biggest by poetair in https://github.com/Yard1/hpbandster-sklearn/pull/8 * Fix issues with resources impacting choice of best model by poetair in https://github.com/Yard1/hpbandster-sklearn/pull/10
New Contributors * poetair made their first contribution in https://github.com/Yard1/hpbandster-sklearn/pull/7
* Fixed an exception during initialization when one of `min_budget`, `max_budget` params was None while other one was not. * The resource (other than `n_samples`) will now be removed from the Configuration Space. * If `refit=True`, the refitted model will use the maximum amount of resource.
2.0.0
Compatibility update for scikit-learn>= 0.24.0. No functionality changes. Not compatible with scikit-learn<0.24.0.
1.1.1
* Fixed an exception during initialization when one of `min_budget`, `max_budget` params was None while other one was not. * The resource (other than `n_samples`) will now be removed from the Configuration Space. * If `refit=True`, the refitted model will use the maximum amount of resource.
1.1.0
* Added support for Pipelines and `TransformedTargetRegressor`. * `error_score` param now defaults to `np.nan` for consistency with scikit-learn tuners.
1.0.1
Fixed an issue with pip installation. Please do not use version 1.0.0.