0 / 0
Model inventory and AI Factsheets
Model inventory and AI Factsheets

Model inventory and AI Factsheets

Use a model inventory to track the lifecycles of machine learning models from training to production. View factsheets for model assets that track lineage events and facilitate efficient ModelOps governance.

Service The Watson Knowledge Catalog, Watson Studio, Watson Machine Learning, and other supplemental services used to track models in an inventory are not available by default. An administrator must install these services on the IBM Cloud Pak for Data platform and they must be installed in the same namespace. To determine whether a service is installed, open the Services catalog and check whether the service is enabled.

Model inventory and governance

Watson Knowledge Catalog provides the capabilities for you to track data science models across the organization. View at a glance which models are in production and which need development or validation. Use the governance features to establish processes to manage the communication flow from data scientists to ModelOps administrators.

Note: A model inventory tracks only the models that you add to entries. You can control which models to track for an organization without tracking samples and other models that are not significant to the organization.

Model inventory

How the Model inventory works

The model inventory is a view in Watson Knowledge Catalog where you can request a new model, then track it through its lifecycle. A typical flow might go as follows:

  1. A business user identifies a need for a machine learning model and creates a model entry to request a new model. The business owner assigns a potential name and states the basic parameters for the requested model.
  2. When the request is saved, a model entry is created in the inventory and the tracking begins. Initially, the entry is in the Awaiting development state because there are no assets to accompany the request.
  3. When a data scientist creates a model for this business case, they track the model from the model details page of the project or deployment space, and associates it with the model entry.
  4. The model entry in the inventory can now be moved to an In progress state and stakeholders can review the assets for the entry, which now include the model.
  5. As the model advances in the lifecycle, the model entry reflects all updates, including deployments and input data assets.
  6. If the data scientist chooses, challenger models can be added to the entry to compare performance.
  7. Validators and other stakeholders can review this and other model entries to ensure compliance with corporate protocols and to view and certify model progress from development to production.

Learn more

Find out about working with a model inventory programmatically, with the IBM_AIGOV_FACTS_CLIENT documentation.

Next steps

Learn about creating and viewing model entries.

Parent topic: Managing AI Lifecycle with ModelOps