tune

Tune submodule for configuring optimisation jobs.

Tuner

Tuner(
    *,
    objective: ObjectiveSpec | list[ObjectiveSpec],
    parameters: list[ParameterSpec],
    num_samples: int,
    mode: Direction | list[Direction] = "max",
    max_concurrent: Optional[int] = None,
    algorithm: Optional[OptunaSpec] = None,
)

A class for running optimisation on Plugboard processes.

Instantiates the Tuner class.

Parameters:

Name	Type	Description	Default
objective	ObjectiveSpec | list[ObjectiveSpec]	The objective(s) to optimise for in the Process.	required
parameters	list[ParameterSpec]	The parameters to optimise over.	required
num_samples	int	The number of trial samples to use for the optimisation.	required
mode	Direction | list[Direction]	The direction of the optimisation.	'max'
max_concurrent	Optional[int]	The maximum number of concurrent trials. Defaults to None, which means that Ray will use its default concurrency of 1 trial per CPU core.	None
algorithm	Optional[OptunaSpec]	Configuration for the underlying Optuna algorithm used for optimisation.	None

is_multi_objective property

is_multi_objective: bool

Returns True if the optimisation is multi-objective.

result_grid property

result_grid: ResultGrid

Returns a [ResultGrid][ray.tune.ResultGrid] summarising the optimisation results.

run

run(
    spec: ProcessSpec,
) -> ray.tune.Result | list[ray.tune.Result]

Run the optimisation job on Ray.

Parameters:

Name	Type	Description	Default
spec	ProcessSpec	The ProcessSpec to optimise.	required

Returns:

Type	Description
Result | list[Result]	Either one or a list of [Result][ray.tune.Result] objects containing the best trial
Result | list[Result]	result. Use the result_grid property to get full trial results.