Databricks Machine Learning Associate Question 478
Select 3You are working on a machine learning project in Databricks where a recommendation system model has been trained. The model is being deployed in a production environment. The team is debating whether to promote code or promote models for faster iterations and seamless updates. Which of the following scenarios justify promoting models instead of promoting code?
- A
The model's hyperparameters have been fine-tuned and you want to deploy the updated model without making any changes to the underlying code.
- B
The data preprocessing logic has changed and requires updates to the feature engineering pipeline.
- C
You need to ensure reproducibility of the model and its dependencies in the production environment.
- D
The machine learning framework or library used in the codebase has been upgraded and requires testing before deployment.
- E
You need to deploy a newly trained version of the model while maintaining the same codebase used in earlier versions.
Show answer and explanation
Correct answers: A, C, E
Explanation
Promoting models is ideal when updates are isolated to the trained model itself, such as fine-tuning or retraining, and the codebase remains unchanged. This approach ensures faster deployment, reproducibility, and version control for model artifacts. On the other hand, promoting code is necessary when changes involve the data pipeline, feature engineering, or dependencies that require updates to the underlying code.
- A. Correct.
Promoting models is preferred when you need to deploy updated model artifacts (e.g., after fine-tuning) without modifying the underlying codebase. This allows faster iteration and avoids unnecessary code changes.
- B. Incorrect.
Promoting code is required when the data preprocessing logic or pipeline changes, as this directly affects the codebase and requires updates to maintain consistency.
- C. Correct.
Promoting models ensures that the exact trained model, along with its dependencies and metadata, is reproducible in production environments. This is critical for maintaining consistency between training and deployment.
- D. Incorrect.
Promoting code is necessary when there are changes to the machine learning framework or library, as these require updates and testing in the codebase to ensure compatibility and functionality.
- E. Correct.
Promoting models is suitable here because it allows you to deploy a new version of the model while keeping the same codebase, ensuring minimal disruption to the development workflow.