bencher.results.optuna_result

Classes

OptunaResult

Functions

_evaluate_over_non_optimized(worker, opt_kwargs, ...)	Evaluate worker across all combinations of non-optimized vars and return mean results.
_aggregate_non_optimized(df, opt_vars, non_opt_vars, ...)	Group DataFrame by optimized vars and average target columns over non-optimized vars.
_study_has_multiple_params(study)	True when the study has >1 trial parameter, making importance meaningful.

Module Contents

bencher.results.optuna_result._evaluate_over_non_optimized(worker, opt_kwargs, non_opt_vars, result_vars)

Evaluate worker across all combinations of non-optimized vars and return mean results.

bencher.results.optuna_result._aggregate_non_optimized(df, opt_vars, non_opt_vars, target_names)

Group DataFrame by optimized vars and average target columns over non-optimized vars.

bencher.results.optuna_result._study_has_multiple_params(study)

True when the study has >1 trial parameter, making importance meaningful.

class bencher.results.optuna_result.OptunaResult(bench_cfg: bencher.bench_cfg.BenchCfg)

Bases: bencher.results.bench_result_base.BenchResultBase

to_optuna_plots(**kwargs) → list[panel.pane.panel]

Create an optuna summary from the benchmark results

Returns:

A list of optuna plot summarising the benchmark process

Return type:

list[pn.pane.panel]

to_optuna_from_results(worker, n_trials=100, extra_results: list[OptunaResult] | None = None, sampler=None)

bench_results_to_optuna_trials(include_meta: bool = True) → list

Convert an xarray dataset to optuna trials so optuna can further optimise or plot.

Parameters:

include_meta (bool) – When True, include all variables (inputs + meta like repeat and over_time) as trial parameters for importance analysis. When False, use only input variables with partition/aggregation via the optimize flag.

Returns:

Optuna trials derived from benchmark results.

Return type:

list[optuna.trial.FrozenTrial]

bench_result_to_study(include_meta: bool) → optuna.Study

get_best_trial_params(canonical=False)

get_pareto_front_params()

collect_optuna_plots(pareto_width: float | None = None, pareto_height: float | None = None) → panel.pane.panel

Use optuna to plot various summaries of the optimisation.

Parameters:

• pareto_width – Optional width for the pareto front plot.

• pareto_height – Optional height for the pareto front plot.

Returns:

A panel with optuna visualisations.

Return type:

pn.pane.panel

deep() → OptunaResult

Return a deep copy of these results

get_best_holomap(name: str | None = None)