Challenge 2 – Create a Build Pipeline in Azure DevOps

< Previous Challenge - Home - Next Challenge >

Introduction

The Build pipeline demonstrates the automation of various stages/tasks involved in building an ML model and creating a container image on top of that model. The stages generally constitute creating a project environment, preparing the data, training the model, evaluating the model, registering/versioning the model, and scoring the model generally by containerizing the model.

There are several ways to create a Build pipeline. The two most common and popular ways are:

• Using a YAML file that represents the entire pipeline.
• Using the classic GUI pipeline and adding tasks sequentially.

Use whichever approach your team is most comfortable with.

We can setup Continuous Integration (CI) trigger for every Build pipeline. The CI pipeline gets triggered every time code is checked in. It publishes an updated Azure Machine Learning pipeline after building the code.

Description

• Make sure you have setup a new project, imported the code, created service connections, and configured Azure ML Workspace for the project.
• Create a Build pipeline.
  • Select the repo that was imported in the previous challenge.
  • Set Agent Pool to Azure Pipelines.
  • Set Agent Specification to ubuntu-18.04.
• Add Build pipeline tasks
  • Add a task to setup environment by using install_environment.sh file in environment_setup/ folder. This will install all the python modules required for the project.
  • Add a task to get Azure ML Workspace connection using Workspace.py in service/code/ folder. This will establish connection to Azure ML workspace by using your workspace details in configuration/config.json file.
  • Add a task to acquire time series transactions data using AcquireData.py in service/code/ folder. This will download and extract the data required to train a forecasting model in the next steps.
  • Add a task to train ARIMA forecasting model using TrainOnLocal.py in service/code/ folder. This will build a model to forecast demand of items from AdventureWorks database.
  • Add a task to evaluate the model performance using EvaluateModel.py in service/code/ folder. This will evaluate how well the model is doing by using evaluation metrics such as R-squared and RMSE(Root mean squared error).
  • Add a task to register the model in Azure ML Model Registry for model versioning using RegisterModel.py in service/code/ folder.
  • Add a task to score the model, to forecast future transactions using CreateScoringImage.py in service/code/ folder. This will create a scoring file.
  • Now you are at a point of creating an artifact for your Release pipeline. An artifact is the deployable component of your model or application. Build Artifact is one of the many artifact types. The following two tasks are required to create Build artifact in your Build pipeline.
    • Use Copy Files task to copy files from $(Build.SourcesDirectory) to $(Build.ArtifactStagingDirectory).
    • Use Publish Artifact task with $(Build.ArtifactStagingDirectory) as path to publish.
• Run the Build pipeline.
• Review Build Outputs - confirm that the model and Azure Container Image have been registered in the Azure ML workspace in respective registries.

Success criteria

• An end-to-end Build pipeline in Azure DevOps.
• Forecasting model registered with the Azure ML Model Registry.

Tips

• Finding the path to where Azure DevOps will copy your build artifact is often the hardest part.
  • You can use the following command in a Bash task to print all environment variables (which is how predefined variables are passed to your pipeline).

    env | sort
    

  • You can use the following command in a Bash task to print a tree of the filesystem of your build agent.

    find $(Pipeline.Workspace) -print | sed -e "s;[^/]*/;|____;g;s;____|; |;g"
    

• Use the predefined variables in Azure DevOps to make your tasks simpler & more robust.
• Make sure you specify the version of Python you want the tasks to use (Python 3.6, there is a task for this)
• Use the Azure CLI task to run the Python scripts since they need to interact with the Azure Machine Learning resource.

Learning resources

• Key concepts for new Azure Pipelines users
• MLOps documentation: Model management, deployment, and monitoring with Azure Machine Learning
• MLOps Reference Architecture