AI-300 Practice Question 62: Serverless Compute Jobs
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 7 · question 2 of 10
A team submits occasional training jobs and wants to avoid creating, sizing and managing compute clusters. They want the lowest administrative effort while jobs still run on demand. What should they do?
Show answer and explanation
Answer:
With serverless compute, a job can be submitted without a compute target. Azure Machine Learning provisions the compute for the job, and the user specifies only resources such as the instance type and instance count. There is no cluster to create or manage (option D).
- A is wrong because this still means creating, sizing and maintaining many clusters, which is the administrative work the team wants to avoid. - B is wrong because a compute instance is a development workstation. Creating and deleting one per job adds work and is not a training target. - C is wrong because an attached cluster must be provisioned, secured and operated by the team, which is more effort than a compute cluster.
Reference: Model training on serverless compute