Machine Learning Designer provides the Update EAS Service (Beta) component. You can connect this component downstream from a model component to update an online service, and submit the pipeline to DataWorks for periodic scheduling to automate online model service updates.
Prerequisites
The model built in Machine Learning Designer has been deployed as an online service and is running properly. For more information, see Deploy a single model as an online service.
Procedure
-
Drag the Update EAS Service (Beta) component onto the canvas. Connect the model output port of the upstream model component to the input port of the Update EAS Service (Beta) component.
The upstream component for the Update EAS Service (Beta) component can also be a model path stored in an OSS bucket. Examples include models output by PMML machine learning algorithms, some computer vision algorithms, some text processing algorithms, and XGBoost training algorithms.
-
Click the Update EAS Service (Beta) component and configure the parameters on the Parameters Settings tab on the right. The RegionId parameter defaults to the current region.
-
EAS service name: The name of the Elastic Algorithm Service (EAS) to update. The service status must be Running.
-
EAS service description json: Leave this parameter blank unless you need to modify other service settings. To do so, refer to the Command reference and enter the required JSON configuration in the text box.
-
-
Right-click the Update EAS Service (Beta) component and click Run Current Node. After the run succeeds, the component creates a new online service version.
-
Use DataWorks to update the online service on a schedule.
To update the online service on a schedule, submit the pipeline to DataWorks for periodic scheduling. For more information, see Use DataWorks for offline scheduling of Machine Learning Designer pipelines.
Related documents
Go to the PAI-EAS online model service page to view service statuses and manage your services. For more information, see Service deployment.