Databricks Machine Learning Professional Question 83
Select 2You are working on a machine learning project in Databricks and have registered a model in the Model Registry. Your team wants to add metadata to the registered model (e.g., a description of its purpose) and to a specific version of the model (e.g., version-specific deployment information). Which of the following methods should you use to accomplish this?
- A
Add a description to the registered model using the
mlflow.register_model()method - B
Use the
mlflow.set_registered_model_tag()method to add tags to the registered model - C
Use the
mlflow.set_model_version_tag()method to add tags to a specific model version - D
Update the metadata of a registered model version using the Databricks UI
- E
Add metadata to the registered model and model version directly in the Databricks Feature Store
Show answer and explanation
Correct answers: B, C
Explanation
To add metadata to a registered model, you use the mlflow.set_registered_model_tag() method, which allows you to attach tags (key-value pairs) to the model. For adding metadata to a specific model version, the mlflow.set_model_version_tag() method is used. These methods are part of the MLflow API, which integrates with the Databricks Model Registry to manage models effectively.
- A. Incorrect.
The
mlflow.register_model()method is used to register a model, not to add metadata like tags or descriptions to a registered model. - B. Correct.
The
mlflow.set_registered_model_tag()method is specifically designed to add tags (key-value metadata) to a registered model. This is the correct approach for adding metadata to a registered model. - C. Correct.
The
mlflow.set_model_version_tag()method is used to add tags (key-value metadata) to a specific version of a registered model. This is the correct approach for adding metadata to a registered model version. - D. Incorrect.
While the Databricks UI allows some interaction with the Model Registry, there is no direct feature for updating metadata for a specific model version in the UI. Metadata updates must be performed using appropriate MLflow methods.
- E. Incorrect.
The Databricks Feature Store is used for managing features, not for managing metadata for registered models or their versions.