Parameters ---------- automl: AutoMLClassifier The fitted automl instance with an ensemble Expects ------- * Perform

Expects ------- * Auto-sklearn can predict/predict_proba and has a model * Each ensemble in the pareto_set can predict/predict_proba

Test that the pipeline item is clonable with `sklearn.clone` Expects ------- * The cloned object has all the same param keys

Test that predict_proba predicts the same as the underlying classifier with predict_proba argument `batches`. Expects -------

Test that predict_proba predicts the same as the underlying classifier with predict_proba argument `batches`, with a sparse dataset E

Test that predict_proba predicts the same as the underlying classifier with predict_proba argument `batches`, for multiclass and multilabel da

Test that predict_proba predicts the same as the underlying classifier with predict_proba argument `batches`, for multiclass and multilabel da

Expects ------- * Should be able to load pickled predictions Note ---- * Not sure this should be supported

Expects ------- * Loading predictions with a given precision should load the expected type

Parameters ---------- dataset: str The name of the dataset task : int The task type of the dataset metrics: Seq

Expects ------- * raise if there was no metric set when calling `_do_dummy_prediction()`

Expects ------- * Should raise an error if the given metric is not applicable to a given task type

Expects ------- * Updating the recorded times should not trigger `was_modified` * Should update the recorded times so `was_modified`

Parameters ---------- memory_allocation: float An invalid memory allocation as a float Expects ------- * Should rais

Parameters ---------- memory_allocation: float An invalid memory allocation as a int Expects ------- * Should raise

Parameters ---------- memory_allocation: Any An invalid memory allocation type Expects ------- * Should raise a Valu

Parameters ---------- dataspec : Dict The spec to make data of expected_shape : Tuple[int, ...] The expected shape o

Test that the default pipeline can be converted to its representation and converted back. Expects ------- * The the S

Expects ------- * All runs should be kept

Expects ------- * Should keep exactly as many models as `max_models` * Should not have any in common between keep and delete

Expects ------- * When deleting runs with multiple objectives, they runs kept should be done so in a roundrobin fashion with respec

Expects ------- * When deleting runs with multiple objectives, they runs kept should be done so in a roundrobin fashion with respec

Expects ------- * Should include runs that were modified, even if they have a loss

Expects ------- * Should include runs that have no loss

Expects ------- * Should keep the expected amount of models * The kept models should be sorted by lowest loss * Should not have a

Expects ------- * Using the same builder as used in the real run should result in the same candidate models for the ensemble. *

Expects ------- * The limits enforced by pynisher should account for the time_buffer and duration to run for + a little bit of over

Fixtures -------- mock_logger: PickleableClientLogger A mock logger to use make_sklearn_dataset : Callable[..., Tuple[np.nda

Makes sure that a given configuration is honored in practice. This method tests that the set hyperparameters actually create objects

Makes sure that a given configuration is honored in practice. This method tests that the set hyperparameters actually create objects

Parameters ---------- tmp_dir: str The temporary directory to use for this test dask_client: dask.distributed.Client

Parameters ---------- tmp_dir: str The temporary directory to use for this test dask_client: dask.distributed.Client

Parameters ---------- budget_type : Union[None, str] The budget type to use Fixtures -------- make_automl : Callable

Expects ------- * With n_jobs set to 1 and no dask client, we default to a SingleThreadedClient with a "fork" _multiprocessing_contex