Azure Machine Learning REST API

REST API reference for Azure Machine Learning model deployment and management

Azure Machine Learning REST API

The Azure Machine Learning REST API enables you to create, manage, and deploy machine learning models using standard HTTP verbs. The API supports workspace management, model registration, online endpoints, and job execution.

Authentication

Azure Machine Learning REST API uses OAuth2 service principal authentication.

Retrieve Authentication Token

1. Get Service Principal Credentials

   Obtain your tenant ID, client ID, and client secret from your Azure service principal.

2. Request Access Token

   curl -X POST https://login.microsoftonline.com/<TENANT-ID>/oauth2/token \
     -d "grant_type=client_credentials&resource=https%3A%2F%2Fmanagement.azure.com%2F&client_id=<CLIENT-ID>&client_secret=<CLIENT-SECRET>"
   
3. Use Token in Requests

   Include the token in the Authorization header:

   Authorization: Bearer <ACCESS-TOKEN>
   

Token Response:

{
  "token_type": "Bearer",
  "expires_in": "3599",
  "access_token": "YOUR-ACCESS-TOKEN"
}

Base URL

https://management.azure.com

Workspaces

List Workspaces

Retrieve all Azure Machine Learning workspaces in a resource group.

Endpoint:

GET /subscriptions/{subscriptionId}/resourceGroups/{resourceGroupName}/providers/Microsoft.MachineLearningServices/workspaces?api-version=2023-10-01

subscriptionId · string · required

Your Azure subscription ID

resourceGroupName · string · required

Name of the resource group

Response:

{
  "value": [
    {
      "id": "/subscriptions/aaaa0a0a-bb1b-cc2c-dd3d-eeeeee4e4e4e/resourceGroups/MyResourceGroup/providers/Microsoft.MachineLearningServices/workspaces/my-workspace",
      "name": "my-workspace",
      "type": "Microsoft.MachineLearningServices/workspaces",
      "location": "centralus",
      "properties": {
        "friendlyName": "My Workspace",
        "creationTime": "2023-01-03T19:56:09.7588299+00:00",
        "workspaceId": "cba12345-abab-abab-abab-ababab123456",
        "discoveryUrl": "https://centralus.experiments.azureml.net/discovery"
      }
    }
  ]
}

Create Workspace

Create a new Azure Machine Learning workspace.

Endpoint:

PUT /subscriptions/{subscriptionId}/resourceGroups/{resourceGroupName}/providers/Microsoft.MachineLearningServices/workspaces/{workspaceName}?api-version=2023-10-01

location · string · required

Azure region for the workspace

properties.storageAccount · string · required

Resource ID of the Azure Storage account

properties.keyVault · string · required

Resource ID of the Azure Key Vault

properties.applicationInsights · string · required

Resource ID of Application Insights

Request:

{
  "location": "eastus",
  "identity": {
    "type": "systemAssigned"
  },
  "properties": {
    "friendlyName": "My ML Workspace",
    "description": "Production workspace",
    "storageAccount": "/subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.Storage/storageAccounts/{storageAccount}",
    "keyVault": "/subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.Keyvault/vaults/{keyVault}",
    "applicationInsights": "/subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.insights/components/{appInsights}"
  }
}

Models

Register Model

Register a trained model in your workspace.

Endpoint:

PUT https://{region}.api.azureml.ms/modelmanagement/v1.0/subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/models/{modelId}?api-version=2023-10-01

modelId · string · required

Unique identifier for the model

name · string · required

Model name

version · integer · required

Model version number

path · string · required

Path to model files in datastore

List Models

Retrieve all registered models in a workspace.

Endpoint:

GET https://{region}.api.azureml.ms/modelmanagement/v1.0/subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/models?api-version=2023-10-01

Online Endpoints

Create Online Endpoint

Deploy a model to an online endpoint for real-time inference.

Endpoint:

PUT /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/onlineEndpoints/{endpointName}?api-version=2024-04-01

name · string · required

Endpoint name (must be unique in the region)

properties.authMode · string

Authentication mode: Key or AADToken

properties.compute · string · required

Compute resource for the endpoint

Request:

{
  "name": "my-endpoint",
  "location": "eastus",
  "properties": {
    "authMode": "Key",
    "description": "Production endpoint"
  }
}

Create Deployment

Create a deployment under an online endpoint.

Endpoint:

PUT /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/onlineEndpoints/{endpoint}/deployments/{deploymentName}?api-version=2024-04-01

model · string · required

Resource ID of the registered model

codeConfiguration · object

Scoring script configuration:

• codeId: Resource ID of the code
• scoringScript: Name of the scoring script

instanceType · string · required

VM size (e.g., Standard_DS3_v2)

instanceCount · integer · required

Number of instances

Invoke Endpoint

Score data using a deployed model.

Endpoint:

POST https://{endpoint-uri}/score

Authorization · string · required

Bearer token or endpoint key

Content-Type · string · required

application/json

Request:

{
  "data": [
    [1, 2, 3, 4, 5],
    [6, 7, 8, 9, 10]
  ]
}

Jobs

Create Training Job

Submit a training job to Azure Machine Learning.

Endpoint:

PUT /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/jobs/{jobId}?api-version=2024-04-01

properties.jobType · string · required

Job type: Command, Sweep, Pipeline

properties.compute · string · required

Compute resource ID

properties.command · string · required

Command to execute

properties.environment · string · required

Environment resource ID

Get Job Status

Retrieve job details and status.

Endpoint:

GET /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/jobs/{jobId}?api-version=2024-04-01

Response:

{
  "id": "job-id",
  "name": "training-job",
  "properties": {
    "status": "Completed",
    "startTime": "2024-01-15T10:30:00Z",
    "endTime": "2024-01-15T11:45:00Z",
    "jobType": "Command"
  }
}

List Jobs

Get all jobs in a workspace.

Endpoint:

GET /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/jobs?api-version=2024-04-01

Compute

Create Compute Resource

Provision a compute cluster for training.

Endpoint:

PUT /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/computes/{computeName}?api-version=2023-10-01

properties.computeType · string · required

Compute type: AmlCompute, ComputeInstance, AKS

properties.properties.vmSize · string · required

VM size (e.g., Standard_D1)

properties.properties.scaleSettings · object · required

Scaling configuration:

• minNodeCount: Minimum nodes
• maxNodeCount: Maximum nodes
• nodeIdleTimeBeforeScaleDown: Idle time before scale down

Request:

{
  "location": "eastus",
  "properties": {
    "computeType": "AmlCompute",
    "properties": {
      "vmSize": "Standard_D1",
      "vmPriority": "Dedicated",
      "scaleSettings": {
        "maxNodeCount": 4,
        "minNodeCount": 0,
        "nodeIdleTimeBeforeScaleDown": "PT30M"
      }
    }
  }
}

List Compute Resources

Endpoint:

GET /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup}/providers/Microsoft.MachineLearningServices/workspaces/{workspace}/computes?api-version=2023-10-01

API Versioning

Always specify the api-version parameter. Current stable version: 2023-10-01

The API version ensures compatibility as the service evolves. Different services may use different versions.

Error Responses

All errors follow Azure standard error format:

{
  "error": {
    "code": "ResourceNotFound",
    "message": "The specified resource was not found",
    "details": []
  }
}

Related Resources

Python SDK

Use the Python SDK for Azure ML

Azure CLI

Command-line interface for Azure ML