Invoking a machine learning model hosted remotely

You use the wizard to configure the invocation of a remote machine learning model.

About this task

Using the configuration wizard, you can select the deployment of a machine learning model hosted on a remote provider service. This deployment is used to generate a pre-filled predictive model. The following remote providers are supported: IBM Watson® Machine Learning, IBM® Open Prediction Service, and Amazon SageMaker.

Alternatively, you can upload a serialized machine learning model directly into the wizard. Both IBM Watson Machine Learning and IBM Open Prediction Service support the upload of Predictive Model Markup Language (PMML) External link opens a new window or tab files. This option is also available with the embedded machine learning provider.

For more information about providers, see Managing local machine learning providers.

Procedure

  1. Open your predictive model and click Configure in the details panel on the right to launch the predictive model configuration wizard.
  2. Select Remote machine learning model and click Next.
  3. Select the provider where the machine learning model that you want to use is stored from the drop-down list.
    The list of deployments that are available from this provider is displayed.
  4. Select a deployment or upload the serialized model you need to make your prediction:
    Option Procedure
    Select deployment
    1. In the list of deployments available, click the arrow Expand arrow next to the machine learning model that you want to use to expand it.
    2. Select a deployment.
    Note: This option is not available with the embedded machine learning provider.
    Upload serialized model
    1. Click the Upload button above the list of deployments available.
    2. Drag your file into the drop target area or click the drop target area and navigate to the file on your local system.
  5. Click Next to define the input parameters that are needed to make the prediction.
    • If an input schema was automatically generated by the provider, you can edit it if necessary. You can update or delete existing parameters, and add new ones.
    • If no input schema was generated, you can:
      • Work in the Form tab and click Add to add parameters.
      • Work in the JSON tab and directly enter a JSON schema.
      • Or use a JSON payload to generate a pre-populated schema by clicking Generate from payload and pasting a JSON payload. The generated schema is editable.
  6. Click Next to test the model invocation.
    This step is optional and allows you to make sure that the model returns the expected results when called. To test your model:
    1. Enter values for each input parameter.
    2. Click Run.
      If the model does not return the expected results, you can go back to the previous step to edit the input schema as required.
    You can use the output of the invocation test to define the output schema in the next step.
  7. Click Next to define the output values of the prediction. To define the output schema, you can:
    • Work in the Form tab and click Add to add values.
    • Work in the JSON tab and directly enter a JSON schema.
    • Use a JSON payload to generate a pre-populated schema by clicking Generate from payload and pasting a JSON payload. The generated schema is editable.
    • Or use the output of the invocation test to generate a pre-populated schema. The generated schema is editable. This option is available only if you tested the model invocation.
  8. Click Apply.

What to do next

You can now map the input and output data from the machine learning model to data types available in your decision service.