共用方式為

MonitorDefinition 類別

  • 參考

注意

這是實驗性類別，可以隨時變更。 請參閱 https://aka.ms/azuremlexperimental 以取得詳細資訊。

監視定義

繼承
azure.ai.ml.entities._mixins.RestTranslatableMixin
MonitorDefinition

建構函式

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

Keyword-Only Parameters

compute
SparkResourceConfiguration

要與監視器相關聯的 Spark 資源組態

monitoring_target
Optional[MonitoringTarget]

與受監視之模型或部署相關聯的 ARM 識別碼物件。

monitoring_signals
Optional[Dict[str, Union[DataDriftSignal , DataQualitySignal, PredictionDriftSignal , FeatureAttributionDriftSignal , CustomMonitoringSignal , GenerationSafetyQualitySignal]]]

要監視的訊號字典。 索引鍵是訊號的名稱，而值是 DataSignal 物件。 DataSignal 物件的接受值為 DataDriftSignal、DataQualitySignal、PredictionDriftSignal、FeatureAttributionDriftSignal 和 CustomMonitoringSignal。

alert_notification
Optional[Union[Literal['azmonitoring'], azure.ai.ml.entities.AlertNotification]]

監視器的警示組態。

範例

建立監視器定義。


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

   monitor_definition = MonitorDefinition(
       compute=SparkResourceConfiguration(instance_type="standard_e4s_v3", runtime_version="3.2"),
       monitoring_target=MonitoringTarget(
           ml_task="Classification",
           endpoint_deployment_id="azureml:fraud_detection_endpoint:fraud_detection_deployment",
       ),
       alert_notification=AlertNotification(emails=["abc@example.com", "def@example.com"]),
   )