cotengra.hyperoptimizers.hyper_baytune

Hyper optimization using baytune.

Module Contents

Functions

convert_param_to_baytune(param)	Convert a search subspace to baytune form.
baytune_init_optimizers(self, methods, space[, ...])	Set-up the baytune optimizer(s).
baytune_get_setting(self)	Get a setting to trial from one of the baytune optimizers.
baytune_report_result(self, setting, trial, score)	Report the result of a trial to the baytune optimizers.

Attributes

BTB_TYPE_TO_HYPERPARAM

cotengra.hyperoptimizers.hyper_baytune.BTB_TYPE_TO_HYPERPARAM

cotengra.hyperoptimizers.hyper_baytune.convert_param_to_baytune(param)

Convert a search subspace to baytune form.

cotengra.hyperoptimizers.hyper_baytune.baytune_init_optimizers(self, methods, space, sampler='GP', method_sampler='UCB1', sampler_opts=None)

Set-up the baytune optimizer(s).

Parameters:

• space (dict[str, dict[str, dict]]) – The search space.

• sampler (str, optional) – Which btb parameter fitter to use - default 'GP' means gaussian process. Other options include 'Uniform' and 'GPEi'. See https://hdi-project.github.io/BTB/api/btb.tuning.tuners.html.

• method_sampler (str, optional) – Which btb selector to use - default ‘UCB1’. See https://hdi-project.github.io/BTB/api/btb.selection.html.

• sampler_opts (dict, optional) – Options to supply to btb.

cotengra.hyperoptimizers.hyper_baytune.baytune_get_setting(self)

Get a setting to trial from one of the baytune optimizers.

cotengra.hyperoptimizers.hyper_baytune.baytune_report_result(self, setting, trial, score)

Report the result of a trial to the baytune optimizers.