AutoMLRun Class

Represents an automated ML experiment run in Azure Machine Learning.

The AutoMLRun class can be used to manage a run, check run status, and retrieve run details once an AutoML run is submitted. For more information on working with experiment runs, see the Run class.

Inheritance

Run

AutoMLRun

Constructor

AutoMLRun(experiment, run_id, **kwargs)

Parameters

experiment

Experiment

Required

The experiment associated with the run.

run_id

str

Required

The ID of the run.

Remarks

An AutoMLRun object is returned when you use the submit method of an experiment.

To retrieve a run that has already started, use the following code:

   from azureml.train.automl.run import AutoMLRun
   ws = Workspace.from_config()
   experiment = ws.experiments['my-experiment-name']
   automl_run = AutoMLRun(experiment, run_id = 'AutoML_9fe201fe-89fd-41cc-905f-2f41a5a98883')

Methods

cancel	Cancel an AutoML run. Return True if the AutoML run was canceled successfully.
cancel_iteration	Cancel a particular child run.
complete	Complete an AutoML Run.
continue_experiment	Continue an existing AutoML experiment.
fail	Fail an AutoML Run. Optionally set the Error property of the run with a message or exception passed to error_details.
get_best_child	Return the child run with the best score for this AutoML Run.
get_guardrails	Print and return detailed results from running Guardrail verification.
get_output	Return the run with the corresponding best pipeline that has already been tested. If no input parameters are provided, get_output returns the best pipeline according to the primary metric. Alternatively, you can use either the iteration or metric parameter to retrieve a particular iteration or the best run per provided metric, respectively.
get_run_sdk_dependencies	Get the SDK run dependencies for a given run.
pause	Return True if the AutoML run was paused successfully. This method is not implemented.
register_model	Register the model with AzureML ACI service.
resume	Return True if the AutoML run was resumed successfully. This method is not implemented.
retry	Return True if the AutoML run was retried successfully. This method is not implemented.
summary	Get a table containing a summary of algorithms attempted and their scores.
wait_for_completion	Wait for the completion of this run. Returns the status object after the wait.

cancel

Cancel an AutoML run.

Return True if the AutoML run was canceled successfully.

cancel()

Returns

None

cancel_iteration

Cancel a particular child run.

cancel_iteration(iteration)

Parameters

iteration

int

Required

The iteration to cancel.

Returns

None

complete

Complete an AutoML Run.

complete(**kwargs)

Returns

None

continue_experiment

Continue an existing AutoML experiment.

continue_experiment(X=None, y=None, sample_weight=None, X_valid=None, y_valid=None, sample_weight_valid=None, data=None, label=None, columns=None, cv_splits_indices=None, spark_context=None, experiment_timeout_hours=None, experiment_exit_score=None, iterations=None, show_output=False, training_data=None, validation_data=None, **kwargs)

Parameters

X

DataFrame or ndarray or <xref:azureml.dataprep.Dataflow>

default value: None

Training features.

y

DataFrame or ndarray or <xref:azureml.dataprep.Dataflow>

default value: None

Training labels.

sample_weight

DataFrame or ndarray or <xref:azureml.dataprep.Dataflow>

default value: None

Sample weights for training data.

X_valid

DataFrame or ndarray or <xref:azureml.dataprep.Dataflow>

default value: None

Validation features.

y_valid

DataFrame or ndarray or <xref:azureml.dataprep.Dataflow>

default value: None

Validation labels.

sample_weight_valid

DataFrame or ndarray or <xref:azureml.dataprep.Dataflow>

default value: None

validation set sample weights.

data

DataFrame

default value: None

Training features and label.

label

str

default value: None

Label column in data.

columns

list(str)

default value: None

A list of allowed columns in the data to use as features.

cv_splits_indices

ndarray

default value: None

Indices where to split training data for cross validation. Each row is a separate cross fold and within each crossfold, provide 2 arrays, the first with the indices for samples to use for training data and the second with the indices to use for validation data. i.e [[t1, v1], [t2, v2], ...] where t1 is the training indices for the first cross fold and v1 is the validation indices for the first cross fold.

spark_context

<xref:SparkContext>

default value: None

Spark context, only applicable when used inside azure databricks/spark environment.

experiment_timeout_hours

float

default value: None

How many additional hours to run this experiment for.

experiment_exit_score

int

default value: None

If specified indicates that the experiment is terminated when this value is reached.

iterations

int

default value: None

How many additional iterations to run for this experiment.

show_output

bool

default value: False

Flag indicating whether to print output to console.

training_data

<xref:azureml.dataprep.Dataflow> or DataFrame

default value: None

Input training data.

validation_data

<xref:azureml.dataprep.Dataflow> or DataFrame

default value: None

Validation data.

Returns

The AutoML parent run.

Return type

AutoMLRun

Exceptions

ConfigException

InvalidRunState

fail

Fail an AutoML Run.

Optionally set the Error property of the run with a message or exception passed to error_details.

fail(error_details=None, error_code=None, _set_status=True, **kwargs)

Parameters

error_details

str or BaseException

default value: None

Optional details of the error.

error_code

str

default value: None

Optional error code of the error for the error classification.

_set_status

bool

default value: True

Indicates whether to send the status event for tracking.

get_best_child

Return the child run with the best score for this AutoML Run.

get_best_child(metric: Optional[str] = None, onnx_compatible: bool = False, **kwargs: Any) -> azureml.core.run.Run

Parameters

metric

str

default value: None

The metric to use to when selecting the best run to return. Defaults to the primary metric.

onnx_compatible

default value: False

Whether to only return runs that generated onnx models.

kwargs

Required

Returns

AutoML Child Run.

get_guardrails

Print and return detailed results from running Guardrail verification.

get_guardrails(to_console: bool = True) -> Dict[str, Any]

Parameters

to_console

bool

default value: True

Indicates whether to write the verification results to the console.

Returns

A dictionary of verifier results.

Return type

dict

Exceptions

UserErrorException

get_output

Return the run with the corresponding best pipeline that has already been tested.

If no input parameters are provided, get_output returns the best pipeline according to the primary metric. Alternatively, you can use either the iteration or metric parameter to retrieve a particular iteration or the best run per provided metric, respectively.

get_output(iteration: Optional[int] = None, metric: Optional[str] = None, return_onnx_model: bool = False, return_split_onnx_model: Optional[azureml.automl.core.onnx_convert.onnx_convert_constants.SplitOnnxModelName] = None, **kwargs: Any) -> Tuple[azureml.core.run.Run, Any]

Parameters

iteration

int

default value: None

The iteration number of the corresponding run and fitted model to return.

metric

str

default value: None

The metric to use to when selecting the best run and fitted model to return.

return_onnx_model

bool

default value: False

This method will return the converted ONNX model if the enable_onnx_compatible_models parameter was set to True in the AutoMLConfig object.

return_split_onnx_model

SplitOnnxModelName

default value: None

The type of the split onnx model to return

Returns

The run, the corresponding fitted model.

Return type

Run, <xref:Model>

Exceptions

ConfigException

OnnxConvertException

NotFoundException

Remarks

If you'd like to inspect the preprocessor(s) and algorithm (estimator) used, you can do so through Model.steps, similar to sklearn.pipeline.Pipeline.steps. For instance, the code below shows how to retrieve the estimator.

   best_run, model = parent_run.get_output()
   estimator = model.steps[-1]

get_run_sdk_dependencies

Get the SDK run dependencies for a given run.

get_run_sdk_dependencies(iteration=None, check_versions=True, **kwargs)

Parameters

iteration

int

default value: None

The iteration number of the fitted run to be retrieved. If None, retrieve the parent environment.

check_versions

bool

default value: True

If True, check the versions with current environment. If False, pass.

Returns

The dictionary of dependencies retrieved from RunHistory.

Return type

dict

Exceptions

ConfigException

pause

Return True if the AutoML run was paused successfully.

This method is not implemented.

pause()

Exceptions

NotImplementedError

register_model

Register the model with AzureML ACI service.

register_model(model_name=None, description=None, tags=None, iteration=None, metric=None)

Parameters

model_name

str

default value: None

The name of the model being deployed.

description

str

default value: None

The description for the model being deployed.

tags

dict

default value: None

Tags for the model being deployed.

iteration

int

default value: None

Override for which model to deploy. Deploys the model for a given iteration.

metric

str

default value: None

Override for which model to deploy. Deploys the best model for a different metric.

Returns

The registered model object.

Return type

<xref:Model>

resume

Return True if the AutoML run was resumed successfully.

This method is not implemented.

resume()

Exceptions

NotImplementedError:

retry

Return True if the AutoML run was retried successfully.

This method is not implemented.

retry()

Exceptions

NotImplementedError

summary

Get a table containing a summary of algorithms attempted and their scores.

summary()

Returns

Pandas DataFrame containing AutoML model statistics.

Return type

DataFrame

wait_for_completion

Wait for the completion of this run.

Returns the status object after the wait.

wait_for_completion(show_output=False, wait_post_processing=False)

Parameters

show_output

bool

default value: False

Indicates whether to show the run output on sys.stdout.

wait_post_processing

bool

default value: False

Indicates whether to wait for the post processing to complete after the run completes.

Returns

The status object.

Return type

dict

Exceptions

ExperimentExecutionException

Attributes

run_id

Return the run ID of the current run.

Returns

The run ID of the current run.

Return type

str