Share via

MonitorDefinition Class

  • Reference

Monitor definition

Constructor

MonitorDefinition(*, compute: ServerlessSparkCompute, monitoring_target: MonitoringTarget | None = None, monitoring_signals: Dict[str, DataDriftSignal | DataQualitySignal | PredictionDriftSignal | FeatureAttributionDriftSignal | CustomMonitoringSignal | GenerationSafetyQualitySignal | GenerationTokenStatisticsSignal] = None, alert_notification: Literal['azmonitoring'] | AlertNotification | None = None)

Keyword-Only Parameters

Name Description
compute
SparkResourceConfiguration

The Spark resource configuration to be associated with the monitor
monitoring_target
Optional[MonitoringTarget]

The ARM ID object associated with the model or deployment that is being monitored.

Default value: None
monitoring_signals
Optional[Dict[str, Union[DataDriftSignal , DataQualitySignal, PredictionDriftSignal , FeatureAttributionDriftSignal , CustomMonitoringSignal , GenerationSafetyQualitySignal , GenerationTokenStatisticsSignal , ModelPerformanceSignal]]]

The dictionary of signals to monitor. The key is the name of the signal and the value is the DataSignal object. Accepted values for the DataSignal objects are DataDriftSignal, DataQualitySignal, PredictionDriftSignal, FeatureAttributionDriftSignal, and CustomMonitoringSignal.

Default value: None
alert_notification
Optional[Union[Literal['azmonitoring'], azure.ai.ml.entities.AlertNotification]]

The alert configuration for the monitor.

Default value: None

Examples

Creating Monitor definition.


   from azure.ai.ml.entities import (
       AlertNotification,
       MonitorDefinition,
       MonitoringTarget,
       SparkResourceConfiguration,
   )

   monitor_definition = MonitorDefinition(
       compute=SparkResourceConfiguration(instance_type="standard_e4s_v3", runtime_version="3.4"),
       monitoring_target=MonitoringTarget(
           ml_task="Classification",
           endpoint_deployment_id="azureml:fraud_detection_endpoint:fraud_detection_deployment",
       ),
       alert_notification=AlertNotification(emails=["[email protected]", "[email protected]"]),
   )