Databricks ML Associate Practice Question 14: UC Registry URI
Databricks Certified Machine Learning Associate · Test 2 · question 4 of 10
A data scientist finishes a training run and calls:
mlflow.register_model("runs:/<run_id>/model", "churn_model")
The call succeeds, but the model appears in the workspace model registry rather than under the Unity Catalog schema ml.churn, where governance requires all models to live.
What should they change?
Show answer and explanation
Answer:
Two things route a registration to Unity Catalog. mlflow.set_registry_uri("databricks-uc") points the MLflow client at the Unity Catalog registry instead of the workspace registry, and the model name must be the three-level catalog.schema.model. The original call gave a bare name and left the registry URI at its workspace default, so MLflow did exactly what it was asked and created a workspace-registry model. Registering then requires CREATE MODEL on the target schema and USE CATALOG / USE SCHEMA on its parents.
Option A grants a privilege on the wrong registry. The call did not fail for lack of permission, it succeeded, so no grant changes where the model lands.
Option B misunderstands the namespace. A model's catalog and schema are part of its identity in Unity Catalog, fixed at registration; tags are descriptive metadata and cannot relocate anything.
Option D assumes a move operation that does not exist. A workspace-registry model is not a Unity Catalog object, so Catalog Explorer cannot drag it into a schema; the version has to be registered into Unity Catalog instead.
Reference: Manage model lifecycle in Unity Catalog