A multi-agency grant project needs experiment pipeline definitions kept separate from production definitions, with required reviews before production changes. Which source-control practice should they use?
Select an answer to reveal the explanation.
Short Explanation
If you're keeping experiment and production YAML in a single main branch, you'll have no real boundary even with pull request reviews, so lab churn can touch live definitions. Tags won't give you separate lines of development for required reviews, and forked repos split your shared history so you'll struggle to enforce consistent production gates. Separate Git branches plus required reviews keep grant partners from shipping a lab tweak into live scoring.
Full Explanation
Maintaining separate Git branches for experiment versus production pipeline definitions with required reviews keeps multi-agency grant work auditable. Experiment branches absorb rapid iteration on Azure Machine Learning pipelines; production branches accept changes only through reviewed merges. Required reviews enforce cross-agency accountability before definitions that spend grant budget or affect shared services go live.
Keeping experiment and production pipeline YAML in a single main branch with required pull request reviews fails because it leaves both definitions on the same branch with no isolation boundary; reviews alone do not separate rapid experiment churn from production, so experiment edits can still affect live definitions. Maintaining separate Git tags for experiment versus production pipeline versions with required approvals fails because tags are static version markers, not development branches with enforceable branch policies; they do not isolate ongoing experiment work or require reviews before production changes merge. Maintaining separate forked repositories for experiment versus production pipelines with required review policies fails because forks fragment shared history and governance across repos, making consistent reviews and reconciliation harder than separate branches in one repository.
Exam caveat: Domain 1 source-control guidance separates experimental and production ML assets and expects review gates; single-main-branch pull request schemes, separate version tags with approvals, and separate forked repos fail when the stem demands separation with required reviews.
Operational check: attempt to push directly to the production pipeline branch and confirm rejection; merge an experiment change via PR with required reviewers; verify the production deploy pipeline reads only the protected branch tip.
Prefer to study offline? Printable PDF study book + test engine ($7.50, unofficial). These questions stay free here.