Skip to navigation
Data Lab

Feature Transform Services

Run the transformation service needed by feature-engineered models.

View as Markdown

A Feature Transform Service supplies the transformations required by feature-engineered models at serving time. Prepare a running service when the deployment form requests one, then keep it available while dependent resources use it.

Before you begin

Check whether the model you are deploying uses feature engineering. Look for an existing running service before creating another one. The deployed-model workflow prompts for a service when one is required.

Create a service

  1. Open Data Lab > Feature Transform Services and select the create control.
  2. Enter a Service name between 3 and 50 characters.
  3. Select Create service. Creation also starts the service and can take up to two minutes.
  4. When the details page opens, inspect Status. If it is STARTING, refresh until the service is RUNNING or an error is reported.
  5. Return to the deployment form and select the running service when prompted.

Open the service details and inspect its related resources before changing its lifecycle. These references help identify what depends on the service. Retain the service ID alongside the deployment when investigating transformation or prediction errors.

Start or stop a service

Use Start service for a stopped or failed service, then refresh its status. Use Stop service when the service is no longer needed by active work and review the confirmation. The server checks whether it is still in use before stopping it.

Stopping a service allows it to be started again later. Keep required services running while dependent deployments use them.

Delete a service

Select Delete service and review the confirmation. Deletion is permanent and terminates a running service. Services with related resources cannot be deleted; inspect the relationships and resolve the dependencies before trying again.

Next steps

Return to Deploy models and select the running service.