Workspace Model Registry with Unity Catalog

[sachinw]

if the "default catalog for the workspace" is to Unity Catalog, how can we access a model from the workspace model registry?

I have already tried  mlflow.set_tracking_uri("databricks") but still try to find catalog in UC.

[Kaniz · Accepted solution]

Hi @sachinw, If your workspace is enabled for Unity Catalog, the default behaviour of the MLflow Python client is.... However, if you want to access a model from the Workspace Model Registry, you need to explicitly tar.... 

Here is how you can do it:

import mlflow mlflow.set_registry_uri("databricks")

This will direct MLflow to use the Workspace Model Registry instead of the Unity Catalog.

The mlflow.set_tracking_uri("databricks") command you mentioned is used to set the tracking URI to a.... This is different from setting the registry URI, which determines where your models are registered.

Please note that the Workspace Model Registry will be deprecated in the future, and it’s recommended.... 

If you want to upgrade to models in Unity Catalog, you can configure the client to access models in ...:

import mlflow mlflow.set_registry_uri("databricks-uc")

I hope this helps! Let me know if you have any other questions. 😊