Standard

A user specifies the hyper-parameter search space.
hpopt create a new HpOpt object with the user-provided configurations.
hpopt tries to find the hyper-parameters that maximizes or minimizes the metric score by training a model with various combinations of hyper-parameters.
The best hyper-parameters is selected and is used by the following full fine-tuning phase.