Skip to content

Latest commit

 

History

History
120 lines (60 loc) · 6.22 KB

README.md

File metadata and controls

120 lines (60 loc) · 6.22 KB
Raw

Compute Resources

Deploy a trained model as a webservice

In previous lessons, we spent much time talking about training a machine learning model, which is a multi-step process involving data preparation, feature engineering, training, evaluation, and model selection. The model training process can be very compute-intensive, with training times spanning across many hours, days, or weeks depending on the amount of data, type of algorithm used, and other factors. A trained model, on the other hand, is used to make decisions on new data quickly. In other words, it infers things about new data it is given based on its training. Making these decisions on new data on-demand is called real-time inferencing.

Overview

In this lab, you learn how to deploy a trained model that can be used as a webservice, hosted on an Azure Kubernetes Service (AKS) cluster. This process is what enables you to use your model for real-time inferencing.

The Azure Machine Learning designer simplifies the process by enabling you to train and deploy your model without writing any code.

Exercise 1: Open a sample training pipeline

Task 1: Open the pipeline authoring editor

  1. In Azure portal, open the available machine learning workspace.

  2. Select Launch now under the Try the new Azure Machine Learning studio message.

    Launch Azure Machine Learning studio.

  3. When you first launch the studio, you may need to set the directory and subscription. If so, you will see this screen:

    Launch Azure Machine Learning studio.

    For the directory, select Udacity and for the subscription, select Azure Sponsorship. For the machine learning workspace, you may see multiple options listed. Select any of these (it doesn't matter which) and then click Get started.

  4. From the studio, select Designer in the left-hand menu. Next, select Sample 1: Regression - Automobile Price Prediction (Basic) under the New pipeline section. This will open a visual pipeline authoring editor.

    The Sample 1 pipeline is selected.

Task 2: Setup the compute target

  1. In the settings panel on the right, select Select compute target.

    The select compute target link is highlighted.

  2. In the Set up compute target editor, select the existing compute target, then select Save.

Note: If you are facing difficulties in accessing pop-up windows or buttons in the user interface, please refer to the Help section in the lab environment.

Select the existing compute target, then select Save.

Task 3: Create a new experiment and submit the pipeline

  1. Select Submit to open the Set up pipline run editor.

    The Submit button is highlighted.

    Please note that the button name in the UI is changed from Run to Submit.

  2. In the Setup pipeline run editor, select Experiment, Create new and provide New experiment name: designer-run, and then select Submit.

    The dialog is displayed with the previously described values.

  3. Wait for the pipeline run to complete. It will take around 10 minutes to complete the run.

Exercise 2: Real-time inference pipeline

Task 1: Create pipeline

  1. Select Create inference pipeline, then select Real-time inference pipeline from the list to create a new inference pipeline.

    Select the real-time inference pipeline option.

Task 2: Submit the pipeline

  1. Select Submit to open the Set up pipeline run editor.

    Select the Submit button.

    Please note that the button name in the UI is changed from Run to Submit.

  2. In the Setup pipeline run editor, select Select existing, then select the experiment you created in an earlier step: designer-run. Select Submit to start the pipeline.

    Select your previous experiment.

  3. Wait for pipeline run to complete. It will take around 7 minutes to complete the run.

Exercise 3: Deploy web service on Azure Kubernetes Service compute

Task 1: Deploy the web service

  1. After the inference pipeline run is finished, select Deploy to open the Set up real-time endpoint editor.

    Select Deploy to open the editor.

  2. In the Set up real-time endpoint editor, select your existing compute target, then select Deploy.

    The dialog shows the selected compute target.

  3. Wait for the deployment to complete. The status of the deployment can be observed above the Pipeline Authoring Editor.

    Deployment completed dialog is shown.

Task 2: Review deployed web service

  1. To view the deployed web service, select the Endpoints section in your Azure Portal Workspace.

  2. Select the deployed web service: sample-1-regression---automobile to open the deployment details page.

    The deployed service is highlighted within the Endpoints blade.

    Note: you have to select the text of the service name to open the deployment details page

Task 3: Review how to consume the deployed web service

  1. Select the Consume tab to observe the following information:

    1. Basic consumption info displays the REST endpoint, Primary key, and Secondary key.

    2. Consumption option shows code samples in C#, Python, and R on how to call the endpoint to consume the webservice.

    The Consume tab is displayed.

Next Steps

Congratulations! You have just learned how to train and deploy a model to an Azure Kubernetes Service (AKS) cluster for real-time inferencing. You can now return to the Udacity portal to continue with the lesson.