MLA-C02 practice question: Model and prompt versioning
Deployment and Orchestration of ML and AI Workflows A free question from our original MLA-C02 practice exam. Try it, then open the answer to see why each option is right or wrong.
A regional bank deploys its credit models through a CI/CD pipeline. Compliance requires a named reviewer to approve each model version before it can reach the production endpoint, and the deployment must always use the version that was approved last.
The team wants the approval record and the model versions to stay in one managed place, so auditors can see which version was approved and when.
Which approach meets these requirements?
Choose one answer.
AA. Store each model artifact in a versioned S3 prefix, and have the deployment pipeline deploy whichever artifact in the folder is the newest.
BB. Register each model version in a model package group with approval status pending, and deploy only versions whose status is approved.
CC. Record the approval in a spreadsheet, and have the deployment pipeline read the version number from the spreadsheet each time.
DD. Give the production endpoint a tag with the approved version, and have the reviewer edit the tag after each approval.
Show the answer and explanation
Answer: B
A. Wrong: The newest file is not necessarily the approved one, and there is no approval record. The stem requires an approval step tied to the version.
B. Right: Correct. The Model Registry keeps model versions in a model package group with an approval status, so the pipeline can deploy only approved versions and auditors can see the approval history. The registry is the managed source of approved models. Learn more: Model Registration Deployment with Model Registry (opens in a new tab)
C. Wrong: A spreadsheet is outside the managed service and has no versioned link to the models. The option does not give the audit trail that the stem requires.
D. Wrong: Tags describe resources and do not record model approval status or deploy only approved versions. The option does not enforce the approval.
Explanation
The SageMaker AI Model Registry holds versions in a model package group with an approval status, and a pipeline can deploy only approved versions. This gives the audit trail compliance needs. Newest files, spreadsheets and tags do not enforce or record approval.
Want more on Model and prompt versioning? The 5 full-length practice sets have 7 more Model and prompt versioning questions, 325 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets →