Databricks Machine Learning Associate Question 477
Select 3You are working on a machine learning project where multiple teams collaborate on developing and deploying models. The project has a stable and reusable feature engineering pipeline, and the models are undergoing frequent updates due to changing business requirements. Which scenario appropriately describes when promoting code is preferred over promoting models, and vice versa?
- A
Promoting code is preferred when the feature engineering pipeline is stable, and new model training runs are necessary for updated results.
- B
Promoting models is preferred when the feature engineering pipeline is unstable, and frequent updates to the data processing logic are required.
- C
Promoting models is preferred when a validated model version needs to be deployed without retraining or changes to the data pipeline.
- D
Promoting code is preferred when the team wants to standardize deployment and retrain models in each environment using the same pipeline.
- E
Promoting code is preferred when the model is static and does not need to be retrained in different environments.
Show answer and explanation
Correct answers: A, C, D
Explanation
Promoting code is preferred when maintaining consistency in data processing and retraining is critical, such as when the feature engineering pipeline is stable or when standardization across environments is required. On the other hand, promoting models is preferred when a validated model version can be reused without retraining or when the model is static, avoiding the need to retrain in different environments.
- A. Correct.
Promoting code is appropriate in this scenario because the feature engineering pipeline is stable, allowing the same pipeline to be used for retraining models in different environments.
- B. Incorrect.
This is incorrect because an unstable feature engineering pipeline would require promoting code, not models, to ensure consistency during updates.
- C. Correct.
Promoting models is appropriate in this scenario because a validated model version can be reused without retraining or modifying the data pipeline, saving time and resources.
- D. Correct.
Promoting code is appropriate here because retraining models using the same pipeline ensures consistent results across environments, which is essential for standardization.
- E. Incorrect.
This is incorrect because promoting code is typically unnecessary for a static model that does not require retraining; promoting the model itself would suffice.