Skip to content

OptunaSampler

Visitor that walks a generated sweep config, samples values with an Optuna trial, and returns a patch you can merge onto a base config. Reach for this when running hyperparameter studies without changing your train function.

Bases: SweepSampler[TBaseModel]

A sampler that uses Optuna for hyperparameter optimization.

Parameters:

  • trial (Trial) –

    The optuna trial to use for hyperparameter optimization.

__init__ 

__init__(trial: Trial) -> None

Initialize the OptunaSampler.

Parameters:

  • trial (Trial) –

    The optuna trial to use for hyperparameter optimization.

sample 

sample(sweep: Sweeper[TBaseModel]) -> TBaseModel

Sample a model from the sweep.

Parameters:

  • sweep (TSweep) –

    The sweep to sample from.

Returns:

  • TBaseModel ( TBaseModel ) –

    The sampled model.