Awesome

GitHub Action for training Machine Learning Models using Azure

Deprecation notice

This Action is deprecated. Instead, consider using the CLI (v2) to manage and interact with Azure Machine Learning jobs (runs) in GitHub Actions.

Important: The CLI (v2) is not recommended for production use while in preview.

Usage

The Azure Machine Learning training action will help you train your models on Azure Machine Learning using GitHub Actions.

Get started today with a free Azure account!

This repository contains a GitHub Action for training machine learning models using Azure Machine Learning in a few different ways, each with different capabilities. To submit a training run, you have to define your python file(s) that should run remotely as well as a config file corresponding to one of the supported methods of training

Dependencies on other GitHub Actions

• Checkout Checkout your Git repository content into GitHub Actions agent.
• aml-workspace This action requires an Azure Machine Learning workspace to be present. You can either create a new one or re-use an existing one using the action.
• aml-compute You can use this action to create a new traininig environment if your workspace doesnt have one already.

Utilize GitHub Actions and Azure Machine Learning to train and deploy a machine learning model

This action is one in a series of actions that can be used to setup an ML Ops process. We suggest getting started with one of our template repositories, which will allow you to create an ML Ops process in less than 5 minutes.

1. Simple template repository: ml-template-azure

   Go to this template and follow the getting started guide to setup an ML Ops process within minutes and learn how to use the Azure Machine Learning GitHub Actions in combination. This template demonstrates a very simple process for training and deploying machine learning models.

2. Advanced template repository: aml-template

   This template demonstrates how the actions can be extended to include the normal pull request approval process and how training and deployment workflows can be split. More enhancements will be added to this template in the future to make it more enterprise ready.

Example workflow for training Machine Learning Models using Azure

name: My Workflow
on: [push, pull_request]
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
    # Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
    - name: Check Out Repository
      id: checkout_repository
      uses: actions/checkout@v2

    # AML Workspace Action
    - uses: Azure/aml-workspace@v1
      id: aml_workspace
      with:
        azure_credentials: ${{ secrets.AZURE_CREDENTIALS }}

    # AML Run Action
    - uses: Azure/aml-run@v1
      id: aml_run
      with:
        # required inputs as secrets
        azure_credentials: ${{ secrets.AZURE_CREDENTIALS }}
        # optional
        parameters_file: "run.json"

Inputs

azure_credentials (Azure Credentials)

Azure credentials are required to connect to your Azure Machine Learning Workspace. These may have been created for an action you are already using in your repository, if so, you can skip the steps below.

Install the Azure CLI on your computer or use the Cloud CLI and execute the following command to generate the required credentials:

# Replace {service-principal-name}, {subscription-id} and {resource-group} with your Azure subscription id and resource group name and any name for your service principle
az ad sp create-for-rbac --name {service-principal-name} \
                         --role contributor \
                         --scopes /subscriptions/{subscription-id}/resourceGroups/{resource-group} \
                         --sdk-auth

This will generate the following JSON output:

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

Add this JSON output as a secret with the name AZURE_CREDENTIALS in your GitHub repository.

parameters_file (Parameter File)

The action tries to load a JSON file in the .cloud/.azure folder in your repository, which specifies details of your Azure Machine Learning Run. By default, the action is looking for a file with the name "run.json". If your JSON file has a different name, you can specify it with this parameter. Note that none of these values are required and in the absence, defaults will be created with a combination of the repo name and branch name.

A sample file can be found in this repository in the folder .cloud/.azure. The JSON file can include the following parameters:

Inputs specific to method of training

• Using a python script (default code/train/run_config.py) which includes a function (default def main(workspace):) that describes your run that you want to submit. If you want to change the default values for the python script, you can specify it with the runconfig_python_file and runconfig_python_function_name parameters. The python script gets the workspace object injected and has to return one of the following objects - Estimator, TensorFlow Estimator, PyTorch Estimator, Scikit Learn Estimator, Chainer Estimator, Pipeline or AutoMLConfig

• Using a runconfig YAML file (default "code/train/run_config.yml"), which describes your Azure Machine Learning Script Run that you want to submit. You can change the default value with the runconfig_yaml_file parameter.

• Using a Pipeline YAML file (default "code/train/pipeline.yml"), which describes your Azure Machine Learning Pipeline that you want to submit. You can change the default value with the pipeline_yaml_file parameter.

Outputs

Other Azure Machine Learning Actions

• aml-workspace - Connects to or creates a new workspace
• aml-compute - Connects to or creates a new compute target in Azure Machine Learning
• aml-run - Submits a ScriptRun, an Estimator or a Pipeline to Azure Machine Learning
• aml-registermodel - Registers a model to Azure Machine Learning
• aml-deploy - Deploys a model and creates an endpoint for the model

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.