Register/ Unregister Model

Register a Model

Registering a model publishes it to the Data Pipeline environment so it can be used for inference when production data is read (for example, via the DS Lab Model Runner).

Supported model types:

• Sklearn (ML & CV)

• Keras (ML & CV)

• PyTorch (ML)

Prerequisites

• The model appears in Data Science Lab > Models.

• If the model is from AutoML, identify the selected candidate you want to register.

Register a Model (from Models)

• Go to Data Science Lab > Models.

• Select an unregistered model in the list.

  • For AutoML experiments containing top candidates, select the target AutoML model via the checkbox in the right-side panel.

• Click Register.

  
• In the Register dialog box, verify the model name (only the selected AutoML model is shown).

• Click Yes to confirm.

• A notification confirms the registration.

  

Verify registration via Filters

• Click Filter.

• In Model Status, choose Registered.

• Click Apply.

• The model now appears under the Registered list.

  

Note: The Register option is also available under the Models section inside a Data Science Notebook. Where it’s used: Registered models are available in the DS Lab Model Runner component within the Data Pipeline module.

Unregister a Model

Unregistering a model removes it from the Data Pipeline environment. The model is no longer available for pipeline inference and moves back to the Unregistered list in Models.

Unregister a Model (from Models)

1. Go to Data Science Lab > Models.

2. Use Filter → Model Status = Registered → Apply to locate the model.

3. Select the registered model.

   • AutoML note: The Models list shows the experiment name; the selected model is displayed in the right panel.

4. Click Unregister.

   
5. In the Unregister dialog box, verify the model name and click Yes.

6. A notification confirms the action.

   

Verify unregistration via Filters

• Click Filter.

• Set Model Status = Unregistered.

• Click Apply.

  
• The model now appears under Unregistered.

  

Effect: The model is removed from the Data Pipeline module, disappears from Registered, and reappears under Unregistered in Models.

Notes & Best Practices

• AutoML: Always confirm which candidate (from the top models) you are registering/unregistering—this is the one reflected in dialogs and action confirmations.

• Pipeline readiness: After registration, validate the model in a non-production pipeline run before promoting to production workflows.

• Version hygiene: If your organization versions models (e.g., via Git or model metadata), record the version committed at registration time for auditability.

Troubleshooting

• Register/Unregister icon disabled: Ensure you have the required permissions and the model is in the correct status (unregistered to register; registered to unregister).

• Model not visible in DS Lab Model Runner: Recheck that the model is registered and you are in the correct project/space; refresh pipeline metadata if applicable.

• AutoML confusion: If the dialog shows an unexpected name, reselect the intended AutoML candidate in the right-side panel and retry.