Use GitHub Actions with Azure Machine Learning

2025-07-31

APPLIES TO: Azure CLI ml extension v2 (current) Python SDK azure-ai-ml v2 (current)

Get started with GitHub Actions to train a model on Azure Machine Learning.

This article teaches you how to create a GitHub Actions workflow that builds and deploys a machine learning model to Azure Machine Learning. You train a scikit-learn linear regression model on the NYC Taxi dataset.

GitHub Actions use a workflow YAML (.yml) file in the /.github/workflows/ path in your repository. This definition contains the various steps and parameters that make up the workflow.

Prerequisites

An Azure Machine Learning workspace. For steps for creating a workspace, see Create the workspace.
The Azure Machine Learning SDK for Python v2. To install the SDK, use the following command:
```
pip install azure-ai-ml azure-identity
```
To update an existing installation of the SDK to the latest version, use the following command:
```
pip install --upgrade azure-ai-ml azure-identity
```
For more information, see Azure Machine Learning Package client library for Python.

A GitHub account. If you don't have one, sign up for free.

Step 1: Get the code

Fork the following repo at GitHub:

https://github.com/azure/azureml-examples

Clone your forked repo locally.

git clone https://github.com/YOUR-USERNAME/azureml-examples

Step 2: Authenticate with Azure

You'll need to first define how to authenticate with Azure. The recommended, more secure option is to sign in with OpenID Connect using a Microsoft Entra application or a user-assigned managed identity. If necessary, you can also use sign in with a service principal and secret. This approach is less secure and not recommended.

Create a service principal with the az ad sp create-for-rbac command in the Azure CLI.

az ad sp create-for-rbac --name "myML" --role contributor \
                            --scopes /subscriptions/<subscription-id>/resourceGroups/<group-name> \
                            --sdk-auth

In the example above, replace the placeholders with your subscription ID, resource group name, and app name. The output is a JSON object with the role assignment credentials that provide access to your App Service app similar to below. Copy this JSON object for later.

  {
    "clientId": "<GUID>",
    "clientSecret": "<GUID>",
    "subscriptionId": "<GUID>",
    "tenantId": "<GUID>",
    (...)
  }

OpenID Connect is an authentication method that uses short-lived tokens. Setting up OpenID Connect with GitHub Actions is more complex process that offers hardened security.

If you don't have an existing application, register a new Active Directory application and service principal that can access resources. Create the Active Directory application.
```
az ad app create --display-name myApp
```
This command will output JSON with an appId that is your client-id. Save the value to use as the AZURE_CLIENT_ID GitHub secret later.

You'll use the objectId value when creating federated credentials with Graph API and reference it as the APPLICATION-OBJECT-ID.
Create a service principal. Replace the $appID with the appId from your JSON output.

This command generates JSON output with a different objectId and will be used in the next step. The new objectId is the assignee-object-id.

Copy the appOwnerTenantId to use as a GitHub secret for AZURE_TENANT_ID later.
```
 az ad sp create --id $appId
```
Create a new role assignment by subscription and object. By default, the role assignment will be tied to your default subscription. Replace $subscriptionId with your subscription ID, $resourceGroupName with your resource group name, and $assigneeObjectId with the generated assignee-object-id. Learn how to manage Azure subscriptions with the Azure CLI.
```
az role assignment create --role contributor --scope /subscriptions/$subscriptionId/resourceGroups/$resourceGroupName --subscription $subscriptionId --assignee-object-id  $assigneeObjectId --assignee-principal-type ServicePrincipal
```
Run the following command to create a new federated identity credential for your active directory application.
- Replace APPLICATION-OBJECT-ID with the objectId (generated while creating app) for your Active Directory application.
- Set a value for CREDENTIAL-NAME to reference later.
- Set the subject. The value of this is defined by GitHub depending on your workflow:
  - Jobs in your GitHub Actions environment: repo:< Organization/Repository >:environment:< Name >
  - For Jobs not tied to an environment, include the ref path for branch/tag based on the ref path used for triggering the workflow: repo:< Organization/Repository >:ref:< ref path>. For example, repo:n-username/ node_express:ref:refs/heads/my-branch or repo:n-username/ node_express:ref:refs/tags/my-tag.
  - For workflows triggered by a pull request event: repo:< Organization/Repository >:pull_request.
```
az rest --method POST --uri 'https://microsoftgraph.chinacloudapi.cn/beta/applications/<APPLICATION-OBJECT-ID>/federatedIdentityCredentials' --body '{"name":"<CREDENTIAL-NAME>","issuer":"https://token.actions.githubusercontent.com","subject":"repo:organization/repository:ref:refs/heads/main","description":"Testing","audiences":["api://AzureADTokenExchange"]}' 
```

To learn how to create a Create an active directory application, service principal, and federated credentials in Azure portal, see Connect GitHub and Azure.

Create secrets

Service principal
OpenID Connect

In GitHub, browse your repository, select Settings > Secrets > Actions. Select New repository secret.
Paste the entire JSON output from the Azure CLI command into the secret's value field. Give the secret the name AZ_CREDS.

You need to provide your application's Client ID, Tenant ID, and Subscription ID to the login action. These values can either be provided directly in the workflow or can be stored in GitHub secrets and referenced in your workflow. Saving the values as GitHub secrets is the more secure option.

In GitHub, browse your repository, select Settings > Secrets > Actions. Select New repository secret.
Create secrets for AZURE_CLIENT_ID, AZURE_TENANT_ID, and AZURE_SUBSCRIPTION_ID. Use these values from your Active Directory application for your GitHub secrets:

GitHub Secret Active Directory Application

AZURE_CLIENT_ID Application (client) ID

AZURE_TENANT_ID Directory (tenant) ID

AZURE_SUBSCRIPTION_ID Subscription ID
Save each secret by selecting Add secret.

GitHub Secret	Active Directory Application
AZURE_CLIENT_ID	Application (client) ID
AZURE_TENANT_ID	Directory (tenant) ID
AZURE_SUBSCRIPTION_ID	Subscription ID

Step 3: Update `setup.sh` to connect to your Azure Machine Learning workspace

You need to update the CLI setup file variables to match your workspace.

In your forked repository, go to azureml-examples/cli/.
Edit setup.sh and update these variables in the file.

Variable Description

GROUP Name of resource group

LOCATION Location of your workspace (example: chinanorth2)

WORKSPACE Name of Azure Machine Learning workspace

Variable	Description
GROUP	Name of resource group
LOCATION	Location of your workspace (example: `chinanorth2`)
WORKSPACE	Name of Azure Machine Learning workspace

Step 4: Update `pipeline.yml` with your compute cluster name

You use a pipeline.yml file to deploy your Azure Machine Learning pipeline. The pipeline is a machine learning pipeline and not a DevOps pipeline. You only need to make this update if you're using a name other than cpu-cluster for your computer cluster name.

In your forked repository, go to azureml-examples/cli/jobs/pipelines/nyc-taxi/pipeline.yml.
Each time you see compute: azureml:cpu-cluster, update the value of cpu-cluster with your compute cluster name. For example, if your cluster is named my-cluster, your new value would be azureml:my-cluster. There are five updates.

Step 5: Run your GitHub Actions workflow

Your workflow authenticates with Azure, sets up the Azure Machine Learning CLI, and uses the CLI to train a model in Azure Machine Learning.

OpenID Connect
Service principal

Your workflow file is made up of a trigger section and jobs:

A trigger starts the workflow in the on section. The workflow runs by default on a cron schedule and when a pull request is made from matching branches and paths. Learn more about events that trigger workflows.
In the jobs section of the workflow, you checkout code and log into Azure with the Azure login action using OpenID Connect.
The jobs section also includes a setup action that installs and sets up the Machine Learning CLI (v2). Once the CLI is installed, the run job action runs your Azure Machine Learning pipeline.yml file to train a model with NYC taxi data.

Enable your workflow

In your forked repository, open .github/workflows/cli-jobs-pipelines-nyc-taxi-pipeline.yml and verify that your workflow looks like this.

name: cli-jobs-pipelines-nyc-taxi-pipeline
on:
  workflow_dispatch:
  schedule:
    - cron: "0 0/4 * * *"
  pull_request:
    branches:
      - main
      - sdk-preview
    paths:
      - cli/jobs/pipelines/nyc-taxi/**
      - .github/workflows/cli-jobs-pipelines-nyc-taxi-pipeline.yml
      - cli/run-pipeline-jobs.sh
      - cli/setup.sh
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
    - name: check out repo
      uses: actions/checkout@v2
    - name: azure login
      uses: azure/login@v2
      with:
          client-id: ${{ secrets.AZURE_CLIENT_ID }}
          tenant-id: ${{ secrets.AZURE_TENANT_ID }}
          subscription-id: ${{ secrets.AZURE_SUBSCRIPTION_ID }}
    - name: setup
      run: bash setup.sh
      working-directory: cli
      continue-on-error: true
    - name: run job
      run: bash -x ../../../run-job.sh pipeline.yml
      working-directory: cli/jobs/pipelines/nyc-taxi

Select View runs.
Enable workflows by selecting I understand my workflows, go ahead and enable them.
Select the cli-jobs-pipelines-nyc-taxi-pipeline workflow and choose to Enable workflow.
Select Run workflow and choose the option to Run workflow now.

Your workflow file is made up of a trigger section and jobs:

A trigger starts the workflow in the on section. The workflow runs by default on a cron schedule and when a pull request is made from matching branches and paths. Learn more about events that trigger workflows.
In the jobs section of the workflow, you checkout code and log into Azure with your service principal secret.
The jobs section also includes a setup action that installs and sets up the Machine Learning CLI (v2). Once the CLI is installed, the run job action runs your Azure Machine Learning pipeline.yml file to train a model with NYC taxi data.

Enable your workflow

In your forked repository, open .github/workflows/cli-jobs-pipelines-nyc-taxi-pipeline.yml and verify that your workflow looks like this.

name: cli-jobs-pipelines-nyc-taxi-pipeline
on:
  workflow_dispatch:
  schedule:
    - cron: "0 0/4 * * *"
  pull_request:
    branches:
      - main
      - sdk-preview
    paths:
      - cli/jobs/pipelines/nyc-taxi/**
      - .github/workflows/cli-jobs-pipelines-nyc-taxi-pipeline.yml
      - cli/run-pipeline-jobs.sh
      - cli/setup.sh
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
    - name: check out repo
      uses: actions/checkout@v2
    - name: azure login
      uses: azure/login@v2
      with:
        creds: ${{secrets.AZURE_CREDENTIALS}}
    - name: setup
      run: bash setup.sh
      working-directory: cli
      continue-on-error: true
    - name: run job
      run: bash -x ../../../run-job.sh pipeline.yml
      working-directory: cli/jobs/pipelines/nyc-taxi

Select View runs.
Enable workflows by selecting I understand my workflows, go ahead and enable them.
Select the cli-jobs-pipelines-nyc-taxi-pipeline workflow and choose to Enable workflow.
Select Run workflow and choose the option to Run workflow now.

Step 6: Verify your workflow run

Open your completed workflow run and verify that the build job ran successfully. You see a green checkmark next to the job.
Open Azure Machine Learning studio and navigate to the nyc-taxi-pipeline-example. Verify that each part of your job (prep, transform, train, predict, score) completed and that you see a green checkmark.

Clean up resources

When your resource group and repository are no longer needed, clean up the resources you deployed by deleting the resource group and your GitHub repository.

Next steps

Create production ML pipelines with Python SDK

Use GitHub Actions with Azure Machine Learning

Prerequisites

Step 1: Get the code

Step 2: Authenticate with Azure

Generate deployment credentials

Create secrets

Step 3: Update setup.sh to connect to your Azure Machine Learning workspace

Step 4: Update pipeline.yml with your compute cluster name

Step 5: Run your GitHub Actions workflow

Enable your workflow

Step 6: Verify your workflow run

Clean up resources

Next steps

Additional resources

Step 3: Update `setup.sh` to connect to your Azure Machine Learning workspace

Step 4: Update `pipeline.yml` with your compute cluster name