AI-300 Practice Question 23: Reusable Pipeline Component
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 3 · question 3 of 10
Three teams each keep their own copy of the same feature-engineering script inside their training pipelines. The copies have drifted apart, and a bug fix one team made never reached the other two. Each pipeline still needs to run the step with its own parameter values and its own input dataset. What should you do?
Show answer and explanation
Answer:
"In Azure Machine Learning, a component represents one reusable step in a pipeline." Microsoft lists share and reuse as the point of them: "components can be easily shared and reused across pipelines, workspaces, and subscriptions. Components built by one team can be discovered and used by another team." Components also require a "well-defined interface (input and output)", which is what lets each pipeline pass its own parameters and dataset into one shared implementation (option B). One fix to the component, published as a new version, reaches every pipeline that references it.
- A is wrong because a mounted folder shares files, not an asset. There is no interface, no version to pin, and a completed pipeline gives you no record of which revision of the script it actually ran. - C is wrong because pulling the step out into a separate job breaks the pipeline graph: you lose the step-level orchestration and lineage, you have to sequence the jobs yourself, and each team is still running its own copy of the script. - D is wrong because, while a wheel does centralize the code, it couples the step to every team's environment image; a fix now requires rebuilding and redeploying environments. The pipeline step still has no declared interface and no independently versioned asset in the workspace.
Reference: What is a component - Azure Machine Learning