Databricks Machine Learning Professional Question 97
Select 3You are a machine learning engineer working for a retail company. Your team uses Databricks Model Registry to manage machine learning models. The team wants to automate the deployment of a newly registered model into a staging environment and trigger a batch scoring job whenever a model version transitions to 'Staging'. Which combination of Databricks features should you use to implement this automation?
- A
Use a Model Registry Webhook configured to trigger on 'Transition to Staging' events
- B
Create a Databricks Job that performs batch scoring and link it with the Webhook
- C
Manually update the model version stage in the Model Registry to trigger the job
- D
Set up an endpoint that listens for Webhook events and triggers the batch scoring job
- E
Use the Databricks AutoML pipeline to automatically deploy the model to staging
Show answer and explanation
Correct answers: A, B, D
Explanation
To automate the model lifecycle in this scenario, you can use Model Registry Webhooks to listen for 'Transition to Staging' events and trigger a batch scoring job. The Webhook can notify an external endpoint or directly invoke Databricks Jobs for further automation. This approach ensures the deployment and scoring process is seamless and automated, eliminating manual intervention.
- A. Correct.
Correct: Model Registry Webhooks can be configured to trigger automated actions when a model version transitions to a specific stage, such as 'Staging'.
- B. Correct.
Correct: Databricks Jobs can be linked to perform batch scoring or other tasks based on triggers like Webhook events.
- C. Incorrect.
Incorrect: Manually updating the model stage is not an automated solution and does not align with the requirement to automate the process.
- D. Correct.
Correct: An endpoint can listen for Webhook events from the Model Registry and trigger the Databricks Job for batch scoring.
- E. Incorrect.
Incorrect: Databricks AutoML simplifies model training and experimentation, but it does not provide automation for deploying models to staging or triggering batch jobs.