Managing models

If you train and deploy multiple versions of models, you need to track various versions and key events for your models. Model lineage provides the means for you to track and manage versions of Watson Machine Learning models.

Model lineage

Use model lineage to track events specific to Watson Machine Learning models. Lineage is the history of the events performed on the Watson Machine Learning asset.

Watch this video to see how to track the lineage of a machine learning model.

Figure 2. Video iconModel Asset Lineage
This video shows you how to track the lineage of a machine learning model.

Lineage events that are tracked include:

  • Model created
  • Model deployed
  • Model re-evaluated
  • Model retrained
  • Model set as an active model

To view lineage for an asset, click the asset name in a project and click the Lineage page. Click a particular event to see details. You can see this type of information about each lineage event:

  • Where: In which project the event occurred.
  • Who: The name of the user who performed the action, unless the action was automated. Automated actions generate lineage events, but don’t show user names.
  • What: A description of the action. Some events show details about the original and updated values.
  • When: The date and time of the event.

For details on reviewing model lineage, see Lineage for Watson Machine Learning assets.