Bitzu

Machine Learning Associate ยท MLflow tracking and the model registry

Registering a model in Unity Catalog

Registering a model with the MLflow Client API / registered_model_name in the Unity Catalog three-level namespace, and the benefits over the workspace model registry (governance, cross-workspace access, lineage).

Key points

  • Use a three-level name catalog.schema.model, either at log time (registered_model_name= in log_model) or afterwards with mlflow.register_model("runs:/<run_id>/model", name).
  • Set mlflow.set_registry_uri("databricks-uc") unless the default is already Unity Catalog.
  • Compared with the workspace registry you get centralized access control, auditing, automatic lineage and access from any workspace on the metastore.

Source: https://docs.databricks.com/aws/en/machine-learning/manage-model-lifecycle/

Practice questions

Single choice

Which name format does a model registered in Unity Catalog use?

  1. model
  2. schema.model
  3. catalog.schema.model
  4. workspace/model/version
Show answer

Answer: C. catalog.schema.model

Unity Catalog models live in the catalog.schema.model namespace. A bare name is the workspace registry style.

Select all that apply

Which are benefits of the Unity Catalog model registry over the workspace registry?

  1. Centralized access control and auditing
  2. Access from any workspace attached to the metastore
  3. Built-in Staging and Production stages
  4. Lineage to upstream tables captured on registration
  5. Registry webhooks for transitions
Show answer

Answer: A. Centralized access control and auditing, B. Access from any workspace attached to the metastore, D. Lineage to upstream tables captured on registration

UC gives centralized access control, lineage and cross-workspace access. Stages and webhooks are not supported in UC, so they are not benefits.

True or false

Models registered in Unity Catalog are visible only in the workspace where they were created.

Show answer

Answer: False

False. UC models are governed at the metastore level, so any workspace attached to it can use them given the right privileges.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More Machine Learning Associate topics in this module