Skip to content

Create Azure Machine Learning Deployment

Actions
Create an Azure Machine Learning Deployment
v0.1.3
Latest
By coding-kitties
Star (1)

Tags

 (1)
deployment

create-azure-machine-learning-deployment

Github Action to create an Azure Machine Learning Online Endpoint

Features:

  • Create an Azure Machine Learning Online Endpoint deployment with the use of a deployment YAML file
  • Update an existing Azure Machine Learning Online Endpoint with use of a deployment YAML file
  • Blue/Green deployment through traffic splitting (Mirroring is also supported)
  • Will automatically scale up/down traffic of new and existing deployments
  • Checks if resource group, workspace and endpoint exist

For other Azure Machine Learning actions check out:

Dependencies on other Github Actions

🚀 Usage

1. Add to Your Workflow

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/[email protected]

      - uses: Azure/login@v1
        with:
          creds: ${{ secrets.AZURE_CREDENTIALS }}

      - name: Create AML Online Endpoint
        uses: coding-kitties/[email protected]
        with:
          deployment_yaml_file_path: "path/to/deployment.yml"
          model_name: "my-model"
          model_version: "1"
          resource_group: "my-resource-group"
          workspace_name: "my-aml-workspace"
          endpoint_name: "my-endpoint"
          registry_name: (optional) "my-registry"
          registry_resource_group: (optional) "my-registry-resource-group"
          traffic: '{ "blue": 80, "green": 20, mirror": {"green": 80} }'

Example deployment of an Azure Machine Learning Workflow with blue/green deployments

This example demonstrates an Azure Machine Learning Deployment with blue/green deployments for different environments. We use various Github Actions to create a complete workflow.

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/[email protected]

      - uses: Azure/login@v1
        with:
          creds: ${{ secrets.AZURE_CREDENTIALS }}

      # Move model into dev registry (Will be skipped if it already exists)
      - name: Register model in registry
        uses: coding-kitties/[email protected]
        with:
          model_name: 'model-name'
          model_version: '1'
          source_registry_name: 'playground-registry'
          source_registry_resource_group: 'my-registry-resource-group'
          destination_registry_name: 'playground-registry'
          destination_registry_resource_group: 'my-registry-resource-group'

      # Create AML Online Endpoint in DEV (Will be skipped if it already exists)
      - name: Create AML Online Endpoint DEV
        uses: coding-kitties/[email protected]
        with:
          endpoint_name: 'dev-endpoint'
          resource_group: 'dev-group'
          workspace_name: 'dev-workspace'

      # Deploy the new green model to DEV
      - name: Create AML Online Endpoint Deployment DEV
        uses: coding-kitties/[email protected]
        with:
          endpoint_name: 'dev-endpoint'
          resource_group: 'dev-group'
          workspace_name: 'dev-workspace'
          deployment_yaml_file_path: 'path/to/deployment.yml'
          model_name: 'model-name'
          model_version: '1'
          traffic: '{ "green": 0, "blue": 100, mirror": {"green": 20} }'

      # Update green deployment traffic in DEV
      - name: Update AML Online Endpoint Deployment traffic
        uses: coding-kitties/[email protected]
        with:
          endpoint_name: 'my-endpoint'
          workspace_name: 'my-workspace'
          resource_group: 'my-resource-group'
          traffic: '{ "green": 100, "blue": 0, mirror": {"green": 0} }'

      - name: Delete AML Online Endpoint Deployment DEV
        uses: coding-kitties/[email protected]
        with:
          endpoint_name: 'dev-endpoint'
          resource_group: 'dev-group'
          workspace_name: 'dev-workspace'
          deployment_name: 'blue'

      # Move model to production registy
      - name: Move model to production registry
        uses: coding-kitties/[email protected]
        with:
          model_name: 'model-name'
          model_version: '1'
          source_registry_name: 'playground-registry'
          source_registry_resource_group: 'my-registry-resource-group'
          destination_registry_name: 'production-registry'
          destination_registry_resource_group: 'my-registry-resource-group'

      # Create AML Online Endpoint in PROD (Will be skipped if it already exists)
      - name: Create AML Online Endpoint PROD
        uses: coding-kitties/[email protected]
        with:
          endpoint_name: 'prod-endpoint'
          resource_group: 'prod-group'
          workspace_name: 'prod-workspace'

      # Deploy the new green model to PROD
      - name: Create AML Online Endpoint Deployment PROD
        uses: coding-kitties/[email protected]
        with:
          endpoint_name: 'prod-endpoint'
          resource_group: 'prod-group'
          workspace_name: 'prod-workspace'
          deployment_yaml_file_path: 'path/to/deployment.yml'
          model_name: 'model-name'
          model_version: '1'
          traffic: '{ "green": 0, "blue": 100, mirror": {"green": 20} }'
å```

Contributors (1)

@MDUYN

Resources

Create Azure Machine Learning Deployment is not certified by GitHub. It is provided by a third-party and is governed by separate terms of service, privacy policy, and support documentation.

About

Create an Azure Machine Learning Deployment
v0.1.3
Latest
By coding-kitties

Tags

 (1)
deployment

Contributors (1)

@MDUYN

Resources

Create Azure Machine Learning Deployment is not certified by GitHub. It is provided by a third-party and is governed by separate terms of service, privacy policy, and support documentation.