hellboyhha / mlmodel-deployment-using-githubaction-on-azuremlstudio

ML model Deployment using GitHub Action on Azure ML Studio

In this tutorial, we will deploy ml model as a real-time managed endpoint in Azure ML Studio using GitHub Action.

Prerequistes

• Azure Subscription ID & Tenant ID
• Azure service principles (Application Client ID & Secret)
• Azure Machine Learning Workspace

What is Azure ML Studio?

Azure Machine Learning Studio is a GUI-based integrated development environment for constructing and operationalizing Machine Learning workflow on Azure. It's a cloud service for accelerating and managing the machine learning project lifecycle. Please see this azure official documentation for more information.

Directory Structure:

Folder [model-requirements/environment]

• includes conda environment yaml file to install dependencies to run ml model

Folder [model-requirements/model]

• includes trained model

Folder [model-requirements/onlinescoring]

• python file that contains the logic about how to run the model and read the input data. please see this for more information.

Folder [.github/workflows]

• includes model_deployment.yaml: yaml file for github action automation

File [model-requirements/sample-request.json]

• test input data to score by using model

File [requirements.txt]

• python dependency to run deployment script from your local or GitHub runner

File [model-deployment-azure.py]

• python file to deploy to azure machine learning studio

Kindly take a moment to review the comments in the files.

GitHub Action Setup

Configure Azure Crendentials as Secrets in GitHub Action

• Secrets for this tutorial.
• You can learn more about GitHub Action Secrets in this documentation.

You can change your model endpoint and deployment name in model-deployment-azure.py

 ....
 # define an endpoint and deployment name
 endpoint_name = <your endpoint name>
 deployment_name = <your deployment name>
 ....

Trigger your GitHub action workflow to deploy model

• Only manual trigger for main branch includes in this repo.
• You can learn more about Triggering GitHub Action Workflow in this documentation.