Databricks Machine Learning Professional Question 87
Select 3A data science team is using Databricks to manage their machine learning models. They want to deploy a model to production but also ensure that they can track ongoing experiments, archive older versions of the model, and rollback if needed. Which of the following model stages in Databricks Model Registry should they use to manage this workflow effectively?
- A
Staging
- B
Production
- C
Archived
- D
Experimentation
Show answer and explanation
Correct answers: A, B, C
Explanation
Databricks Model Registry provides stages like 'Staging', 'Production', and 'Archived' to help manage the lifecycle of machine learning models. 'Staging' is for pre-production testing, 'Production' is for live deployments, and 'Archived' is for storing older versions of models for later use or rollback. The team should use these stages to effectively manage their workflow. 'Experimentation' is not a recognized model stage in Databricks.
- A. Correct.
The 'Staging' stage is used to test the model in a pre-production environment. It is a crucial step for validating the model before moving it to production.
- B. Correct.
The 'Production' stage is used for deploying the model in a live environment where it serves predictions. Models in this stage are actively used by applications.
- C. Correct.
The 'Archived' stage is used to store older versions of models that are no longer active. This allows rollback to previous versions if needed.
- D. Incorrect.
The 'Experimentation' stage is not a valid model stage in Databricks Model Registry. Experimentation is typically handled in the MLflow tracking server, not through model stages.