Databricks Machine Learning Associate Question 491
Single answerYou are managing machine learning models in a Databricks MLflow model registry. Your team has deployed a champion model for a recommendation system, but you recently trained a new challenger model that shows improved performance metrics. How can you promote the challenger model to replace the current champion model using aliases?
- A
Use the MLflow CLI to delete the current 'champion' alias and assign it to the new model version.
- B
Assign the 'champion' alias to the challenger model version directly, which will automatically remove it from the previous model version.
- C
Register the challenger model as a new model and assign the 'champion' alias to it manually.
- D
Promote the challenger model to 'Production', and the 'champion' alias will be automatically updated.
Show answer and explanation
Correct answer: B
Explanation
In MLflow, aliases like 'champion' are used to identify specific model versions. When promoting a challenger model to a champion model, you can directly reassign the 'champion' alias to the new model version. This will automatically remove the alias from the previous champion model version, ensuring that the alias always points to the current champion model.
- A. Incorrect.
This is incorrect because deleting aliases manually is not required. MLflow allows you to reassign aliases directly to a new model version, simplifying the process.
- B. Correct.
This is correct because MLflow aliases like 'champion' can be directly reassigned to a new model version, and this action will automatically remove the alias from the previous model version.
- C. Incorrect.
This is incorrect because registering the challenger model as a new model is unnecessary. You only need to assign the 'champion' alias to the appropriate model version.
- D. Incorrect.
This is incorrect because promoting a model to 'Production' does not automatically update aliases. Aliases must be reassigned explicitly to reflect changes.