Train a model with the Azure Machine Learning Designer

Azure Machine Learning Designer provides a drag and drop interface with which you can define a workflow. You can create a workflow to train a model, testing and comparing multiple algorithms with ease.

In this exercise, you’ll use the Designer to quickly train and compare two classification algorithms.

Before you start

You’ll need an Azure subscription in which you have administrative-level access.

Provision an Azure Machine Learning workspace

An Azure Machine Learning workspace provides a central place for managing all resources and assets you need to train and manage your models. You can interact with the Azure Machine Learning workspace through the studio, Python SDK, and Azure CLI.

You’ll use a Shell script which uses the Azure CLI to provision the workspace and necessary resources. Next, you’ll use the Designer in the Azure Machine Learning studio to train and compare models.

Create the workspace and compute cluster

To create the Azure Machine Learning workspace and a compute cluster, you’ll use the Azure CLI. All necessary commands are grouped in a Shell script for you to execute.

  1. In a browser, open the Azure portal at https://portal.azure.com/, signing in with your Microsoft account.
  2. Select the [>_] (Cloud Shell) button at the top of the page to the right of the search box. This opens a Cloud Shell pane at the bottom of the portal.
  3. Select Bash if asked. The first time you open the cloud shell, you will be asked to choose the type of shell you want to use (Bash or PowerShell).
  4. Check that the correct subscription is specified and that No storage account required is selected. Select Apply.

  5. In the terminal, enter the following commands to clone this repo:

     rm -r azure-ml-labs -f
 git clone https://github.com/MicrosoftLearning/mslearn-azure-ml.git azure-ml-labs

    Use SHIFT + INSERT to paste your copied code into the Cloud Shell.

  6. After the repo has been cloned, enter the following commands to change to the folder for this lab and run the setup.sh script it contains:

     cd azure-ml-labs/Labs/05
 ./setup.sh

    Ignore any (error) messages that say that the extensions were not installed.

  7. Wait for the script to complete - this typically takes around 5-10 minutes.

    Troubleshooting tip: Workspace creation error

    If you receive an error when running the setup script through the CLI, you need to provision the resources manually:

    1. In the Azure portal home page, select + Create a resource.
    2. Search for machine learning and then select Azure Machine Learning. Select Create.
    3. Create a new Azure Machine Learning resource with the following settings:
      • Subscription: Your Azure subscription
      • Resource group: rg-dp100-labs
      • Workspace name: mlw-dp100-labs
      • Region: Select the geographical region closest to you
      • Storage account: Note the default new storage account that will be created for your workspace
      • Key vault: Note the default new key vault that will be created for your workspace
      • Application insights: Note the default new application insights resource that will be created for your workspace
      • Container registry: None (one will be created automatically the first time you deploy a model to a container)
    4. Select Review + create and wait for the workspace and its associated resources to be created - this typically takes around 5 minutes.
    5. Select Go to resource and in its Overview page, select Launch studio. Another tab will open in your browser to open the Azure Machine Learning studio.
    6. Close any pop-ups that appear in the studio.
    7. Within the Azure Machine Learning studio, navigate to the Compute page and select + New under the Compute instances tab.
    8. Give the compute instance a unique name and then select Standard_DS11_v2 as the virtual machine size.
    9. Select Review + create and then select Create.
    10. Next, select the Compute clusters tab and select + New.
    11. Choose the same region as the one where you created your workspace and then select Standard_DS11_v2 as the virtual machine size. Select Next
    12. Give the cluster a unique name and then select Create.
      13. </ol> </details>

Configure a new pipeline

When you’ve created the workspace and necessary compute cluster, you can open the Azure Machine Learning studio and create a training pipeline with the Designer.

  1. In the Azure portal, navigate to the Azure Machine Learning workspace named mlw-dp100-….
  2. Select the Azure Machine Learning workspace, and in its Overview page, select Launch studio. Another tab will open in your browser to open the Azure Machine Learning studio.
  3. Close any pop-ups that appear in the studio.
  4. Within the Azure Machine Learning studio, navigate to the Compute page and verify that the compute cluster you created in the previous section exist. The cluster should be idle and have 0 nodes running.
  5. Navigate to the Designer page.
  6. Select the Custom tab at the top of the page.
  7. Create a new empty pipeline using custom components.
  8. Change the default pipeline name (Pipeline-Created-on-date) to Train-Diabetes-Classifier by selecting the pencil icon on its right.

Create a new pipeline

To train a model, you’ll need data. You can use any data stored in a datastore or use a publicly accessible URL.

  1. In the left menu, select the Data tab.

  2. Drag and drop the diabetes-folder component to the canvas.

    Now that you have your data, you can continue by creating a pipeline using custom components that already exist within the workspace (were created for you during setup).

  3. In the left menu, select the Components tab.
  4. Drag and drop the Remove Empty Rows component on to the canvas, below the diabetes-folder.
  5. Connect the output of the data to the input of the new component.
  6. Drag and drop the Normalize Numerical Columns component on to the canvas, below the Remove Empty Rows.
  7. Connect the output of the previous component to the input of the new component.
  8. Drag and drop the Train a Decision Tree Classifier Model component on to the canvas, below the Normalize Numerical Columns.
  9. Connect the output of the previous component to the input of the new component.
  10. Select Configure & Submit and in the Set up pipeline job page create a new experiment and name it diabetes-designer-pipeline, then select Next.
  11. On the Inputs & Outputs make no changes and select Next.
  12. On the Runtime settings select Compute Cluster, and under the Select Azure ML compute cluster select your aml-cluster.
  13. Select Review + Submit and then select Submit to start the pipeline run.
  14. You can check the status of the run by going to the Pipelines page and selecting the Train-Diabetes-Classifier pipeline.

  15. Wait until all components have successfully completed.

    Submitting the job will initialize the compute cluster. As the compute cluster was idle up until now, it may take some time for the cluster to resize to more than 0 nodes. Once the cluster has resized, it will automatically start running the pipeline.

You’ll be able to track the run of each component. When the pipeline fails, you’ll be able to explore which component failed and why it failed. Error messages will show in the Outputs + logs tab of the job overview.

Train a second model to compare

To compare between algorithms and evaluate which performs better, you can train two models within one pipeline and compare.

  1. Return to the Designer and select the Train-Diabetes-Classifier pipeline draft.
  2. Add the Train a Logistic Regression Classifier Model component to the canvas, next to the other training component.
  3. Connect the output of the Normalize Numerical Columns component to the input of the new training component.
  4. At the top, select Configure & Submit.
  5. On the Basics page, create a new experiment named designer-compare-classification, and run it.
  6. Select Review + Submit and then select Submit to start the pipeline run.
  7. You can check the status of the run by going to the Pipelines page and selecting the Train-Diabetes-Classifier pipeline with the designer-compare-classification experiment.
  8. Wait until all components have successfully completed.
  9. Select Job overview, then select the Metrics tab to review the results for both training components.
  10. Try and determine which model performed better.

Delete Azure resources

When you finish exploring Azure Machine Learning, you should delete the resources you’ve created to avoid unnecessary Azure costs.

  1. Close the Azure Machine Learning studio tab and return to the Azure portal.
  2. In the Azure portal, on the Home page, select Resource groups.
  3. Select the rg-dp100-… resource group.
  4. At the top of the Overview page for your resource group, select Delete resource group.
  5. Enter the resource group name to confirm you want to delete it, and select Delete.