Skip to main content
Version: Cloud

Configure Azure Machine Learning

With the Azure Machine Learning integration, you can manage, monitor, and explain your Azure Machine Learning Deployments in Deeploy.

Two types of Deployments are supported:

  1. A Deeploy-hosted Deployment (with the KServe deployment backend) that deploys models and explainers from your Azure Machine Learning model registry.
  2. An Azure Machine Learning-hosted Deployment that runs in your own Azure Machine Learning workspace and can optionally deploy models and explainers from your Azure Machine Learning Registry.

Configure Azure Machine Learning​

The Azure Machine Learning integration is controlled at the organization level. Admins can set up the integration on the Integrations page in the Admin panel. Select Configure on the Azure Machine Learning card and then Add credentials to set up the integration.

Prerequisites​

You need an Azure Machine Learning workspace to create Deployments.

You also need an Azure Active Directory (AD) app that allows Deeploy to connect to your Azure Machine Learning workspace.

Set up the integration​

You need the following information to set up the Azure Machine Learning integration:

  • Azure tenant ID, Azure subscription ID, and Azure resource group name
    You can find these in the Azure portal.
  • Azure client ID and Azure client secret
    Deeploy needs Azure AD app credentials to access the necessary resources in your Azure account. To get your Azure credentials, create an Azure client secret in your Azure AD app. Your app needs a role with the following minimal permissions:
"permissions": [
    {
        "actions": [
            "Microsoft.MachineLearningServices/workspaces/environments/*",
            "Microsoft.MachineLearningServices/workspaces/onlineEndpoints/*",
            "Microsoft.MachineLearningServices/workspaces/models/*/read",
            "Microsoft.MachineLearningServices/workspaces/read",
            "Microsoft.MachineLearningServices/workspaces/metadata/*"
        ],
        "notActions": [],
        "dataActions": [],
        "notDataActions": []
    }
]

Update Azure Machine Learning credentials​

To update your Azure Machine Learning credentials, see update integration credentials. When you update your Azure Machine Learning credentials, you must provide your client secret.

Get started with Azure Machine Learning for Deployments​

Get started with Azure Machine Learning Deployments by making Azure Machine Learning your default deployment service or by creating individual Azure Machine Learning Deployments. This section explains both options.

For guidance on creating Deployments with Azure Machine Learning as a deployment service, see Creating Azure Machine Learning Deployments.

Set Azure Machine Learning as your default deployment service​

The default deployment service is controlled at the Workspace level. Workspace admins can change the default deployment service to Azure Machine Learning in the Workspace settings. All Deployments use the default deployment service unless you specify otherwise when creating a Deployment.

Create an individual Azure Machine Learning Deployment​

Create a Deployment using Azure Machine Learning when the default deployment service is set to KServe or SageMaker (and vice versa). To change the deployment service for an individual Deployment, clear the Use default deployment service settings option in the Deployment step when creating your Deployment, and select Azure Machine Learning as your deployment service.