lale.lib.rasl.task_graphs module

class lale.lib.rasl.task_graphs.Prio[source]

Bases: ABC

Abstract base class for scheduling priority in task graphs.

arity: int
batch_priority(batch: _Batch) → Any[source]
bottom() → Any[source]
abstract task_priority(task: _Task) → Any[source]
class lale.lib.rasl.task_graphs.PrioBatch[source]

Bases: Prio

Execute tasks from earlier batches first.

arity: int = 6
task_priority(task: _Task) → Any[source]
class lale.lib.rasl.task_graphs.PrioResourceAware[source]

Bases: Prio

Execute tasks with less non-resident data first.

arity: int = 5
task_priority(task: _Task) → Any[source]
class lale.lib.rasl.task_graphs.PrioStep[source]

Bases: Prio

Execute tasks from earlier steps first, like nested-loop algorithm.

arity: int = 6
task_priority(task: _Task) → Any[source]
lale.lib.rasl.task_graphs.cross_val_score(pipeline: TrainablePipeline[TrainableIndividualOp], batches: Iterable[Tuple[Any, Any]], scoring: MetricMonoidFactory, cv, unique_class_labels: List[Union[str, int, float]], max_resident: Optional[int], prio: Prio, same_fold: bool, verbose: int) → List[float][source]

Replacement for sklearn’s cross_val_score function (early interface, subject to change).

lale.lib.rasl.task_graphs.cross_validate(pipeline: TrainablePipeline[TrainableIndividualOp], batches: Iterable[Tuple[Any, Any]], scoring: MetricMonoidFactory, cv, unique_class_labels: List[Union[str, int, float]], max_resident: Optional[int], prio: Prio, same_fold: bool, return_estimator: bool, verbose: int) → Dict[str, Union[List[float], List[TrainedPipeline]]][source]

Replacement for sklearn’s cross_validate function (early interface, subject to change).

lale.lib.rasl.task_graphs.fit_with_batches(pipeline: TrainablePipeline[TrainableIndividualOp], batches_train: Iterable[Tuple[Any, Any]], batches_valid: Optional[List[Tuple[Any, Any]]], scoring: Optional[MetricMonoidFactory], unique_class_labels: List[Union[str, int, float]], max_resident: Optional[int], prio: Prio, partial_transform: Union[bool, str], verbose: int, progress_callback: Optional[Callable[[float, float, int, bool], None]]) → TrainedPipeline[TrainedIndividualOp][source]

Replacement for the fit method on a pipeline (early interface, subject to change).

lale.lib.rasl.task_graphs.is_associative(op: TrainableIndividualOp) → bool[source]

Is the operator pre-trained or does it implement MonoidFactory?

lale.lib.rasl.task_graphs.is_incremental(op: TrainableIndividualOp) → bool[source]

Does the operator have a partial_fit method or is it pre-trained?

lale.lib.rasl.task_graphs.is_pretrained(op: TrainableIndividualOp) → bool[source]

Is the operator frozen-trained or does it lack a fit method?