mlos_bench.optimizers.mlos_core_optimizer module

A wrapper for mlos_core optimizers for mlos_bench.

class mlos_bench.optimizers.mlos_core_optimizer.MlosCoreOptimizer(tunables: TunableGroups, config: dict, global_config: dict | None = None, service: Service | None = None)

Bases: Optimizer

A wrapper class for the mlos_core optimizers.

Attributes:

config_space

Get the tunable parameters of the optimizer as a ConfigurationSpace.

current_iteration

The current number of iterations (suggestions) registered.

max_suggestions

The maximum number of iterations (suggestions) to run.

name

The name of the optimizer.

seed

The random seed for the optimizer.

start_with_defaults

Return True if the optimizer should start with the default values.

supports_preload

Return True if the optimizer supports pre-loading the data from previous experiments.

targets

A dictionary of {target: direction} of optimization targets.

tunable_params

Get the tunable parameters of the optimizer as TunableGroups.

Methods

bulk_register(configs, scores[, status])	Pre-load the optimizer with the bulk data from previous experiments.
get_best_observation()	Get the best observation so far.
not_converged()	Return True if not converged, False otherwise.
register(tunables, status[, score])	Register the observation for the given configuration.
suggest()	Generate the next suggestion.

bulk_register(configs: Sequence[dict], scores: Sequence[Dict[str, int | float | str | None] | None], status: Sequence[Status] | None = None) → bool

Pre-load the optimizer with the bulk data from previous experiments.

Parameters:

configsSequence[dict]

Records of tunable values from other experiments.

scoresSequence[Optional[Dict[str, TunableValue]]]

Benchmark results from experiments that correspond to configs.

statusOptional[Sequence[Status]]

Status of the experiments that correspond to configs.

Returns:

is_not_emptybool

True if there is data to register, false otherwise.

get_best_observation() → Tuple[Dict[str, float], TunableGroups] | Tuple[None, None]

Get the best observation so far.

Returns:

(value, tunables)Tuple[Dict[str, float], TunableGroups]

The best value and the corresponding configuration. (None, None) if no successful observation has been registered yet.

property name: str

The name of the optimizer.

We save this information in mlos_bench storage to track the source of each configuration.

register(tunables: TunableGroups, status: Status, score: Dict[str, int | float | str | None] | None = None) → Dict[str, float] | None

Register the observation for the given configuration.

Parameters:

tunablesTunableGroups

The configuration that has been benchmarked. Usually it’s the same config that the .suggest() method returned.

statusStatus

Final status of the experiment (e.g., SUCCEEDED or FAILED).

scoreOptional[Dict[str, TunableValue]]

A dict with the final benchmark results. None if the experiment was not successful.

Returns:

valueOptional[Dict[str, float]]

Benchmark scores extracted (and possibly transformed) from the dataframe that’s being MINIMIZED.

suggest() → TunableGroups

Generate the next suggestion. Base class’ implementation increments the iteration count and returns the current values of the tunables.

Returns:

tunablesTunableGroups

The next configuration to benchmark. These are the same tunables we pass to the constructor, but with the values set to the next suggestion.