Skip to content
Products
/
/
/
Set Up an Automl Solution
Last updated

Set Up an Automl Solution

To set up an AutoML solution, you can create AutoML workflows via TD Console. For the base package, you need to run two workflows: one configured with the ml_train > operator and the other with the ml_predict > operator. For a custom solution package, you need to create a single workflow configured with the ipynb > operator. The overall AutoML setup process is as shown in the following sections:

Create an AutoML workflow

  1. Open the TD Console.
  2. Navigate to Data Workbench > Workflows.
  3. Select New User-Defined Workflow.
  4. Select or Enter a Project Name.
  5. Enter a Workflow Name.
  6. Select a ML notebook solution in the Workflow Template. For example, Gluon Train, Gluon Predict, or a custom solution operator.
  7. Select New Workflow.
  8. View preview.
    The workflow parameters can be edited through the console in the subsequent steps, without editing the workflow code directly.
  9. Select Save & Commit.

Configure a Secret for an AutoML Workflow

To prepare the AutoML workflow, it is necessary to provide a Master Key as a secret. This is required to enable the AutoML container to access relevant databases. You can set and edit secrets or parameters for an AutoML Workflow from the console, like a typical TD workflow.

  1. Select the Secrets tab within the workflow:
  2. Select the PLUS sign (+) to add a new secret.
  3. In the Name field, enter exactly this text: td.apikey
  4. In the Secret Key field, paste in the Master key. (User API Keys can be found in the “API Keys” section of your user profile.)
  5. Select Create. The key will now be available in the list of secrets.

For general details about configuring workflow secrets, see Setting Workflow Secrets from TD Console and Setting Workflow Secrets from the Command Line.

Configure AutoML Notebook Parameters

  1. Select the AutoML tab and select the notebook listed under Notebook Parameters.

  2. The notebook parameters editor appears.

  3. The Input Table can be selected from existing database tables within your Treasure Data environment.

  4. It is important to specify which columns should be ignored, for example any unique identifiers or timestamps.

  5. To share your model, select the Share Model checkbox as shown. See the section AutoML Model Sharing.

  6. Select Confirm to save the parameter changes.

Previous page
Next page