Deploy an ML Model
Last updated
Last updated
There are multiple ways to deploy a model in Azure Machine Learning. One of the simplest ways is to use the designer to automate the deployment process. Use the following steps to deploy a model as a real-time endpoint:
Run your completed training pipeline at least once.
After the job completes, at the top of the canvas, select Create inference pipeline > Real-time inference pipeline.
The designer converts the training pipeline into a real-time inference pipeline. A similar conversion also occurs in Studio (classic).
In the designer, the conversion step also registers the trained model to your Azure Machine Learning workspace.
Your new pipeline looks like this.
Select Submit to run the real-time inference pipeline, and verify that it runs successfully.
After you verify the inference pipeline, select Deploy.
Enter a name for your endpoint and a compute type.
In the dialog box that appears, you can select from any existing Azure Kubernetes Service (AKS) Compute clusters to deploy your model to. If you don't have an AKS cluster, use the following steps to create one.
Select Compute in the left navigation to go to the Compute page.
On the navigation ribbon, select Kubernetes Clusters > + New
Select the Region and specifications for the compute instance.
Click Create.
After completing the above steps, go back to the inference pipeline and complete the Deploy dialogue box as shown in Step 6. Creation of the endpoint may take a while. The endpoint will be Transitioning while it is being created. Wait until the Deployment state of the endpoint changes from Transitioning to Healthy.
After deployment completes, you can see more details and test your endpoint:
Go to the Endpoints tab.
Select your endpoint.
Select the Test tab.
