AI-300 Practice Question 34: Archive Model Versions
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 4 · question 4 of 10
A workspace has eight registered versions of a fraud model. Versions 1 to 5 are obsolete, and the team wants them to stop appearing when people list models, so nobody deploys one by mistake. One scheduled pipeline still references azureml:fraud:3, and it must keep working until it is migrated next quarter. Which action should the team take for the obsolete versions?
Show answer and explanation
Answer:
In Azure Machine Learning, *"archiving a model hides it from list queries like az ml model list by default. You can continue to reference and use an archived model in your workflows."* You can archive all versions or specific ones (az ml model archive --name fraud --version 3), so the obsolete versions disappear from the default listing while the pipeline that references version 3 continues to run (option D). One detail to remember for later: a new version created under an archived model container is automatically archived as well.
- A is wrong because for models *"you can only update the description and tags properties. All other properties are immutable,"* so a version's artifacts cannot be repointed. A change needs a new version. - B is wrong because tags are only metadata. Removing them neither hides the versions from listings nor prevents anyone from deploying them. - C is wrong because the model name is immutable too, and renaming would break the pipeline's reference to fraud:3 even if it were possible.
Reference: Register and work with models - Azure Machine Learning