CERTIFIED-MACHINE-LEARNING-PROFESSIONAL · Question #13
CERTIFIED-MACHINE-LEARNING-PROFESSIONAL Question #13: Real Exam Question with Answer & Explanation
The correct answer is B: The version number of the model version in Production is necessary to complete this task.. Option B is correct because Databricks MLflow Model Serving invocation endpoints require an explicit version number in the URL - stage names like "Production" cannot be substituted for the integer version in the REST endpoint path. The correct URL format is https://<databricks-in
Question
A machine learning engineer has deployed a model recommender using MLflow Model Serving. They now want to query the version of that model that is in the Production stage of the MLflow Model Registry. Which of the following model URIs can be used to query the described model version?
Options
- Ahttps://<databricks-instance>/model-serving/recommender/Production/invocations
- BThe version number of the model version in Production is necessary to complete this task.
- Chttps://<databricks-instance>/model/recommender/stage-production/invocations
- Dhttps://<databricks-instance>/model-serving/recommender/stage-production/invocations
- Ehttps://<databricks-instance>/model/recommender/Production/invocations
Explanation
Option B is correct because Databricks MLflow Model Serving invocation endpoints require an explicit version number in the URL - stage names like "Production" cannot be substituted for the integer version in the REST endpoint path. The correct URL format is https://<databricks-instance>/model/<model-name>/<version-number>/invocations, so you must first look up which version is currently tagged as Production in the Model Registry, then use that number.
Why the distractors fail:
- A & D use
/model-serving/in the path - that's not the valid invocation path for querying a registered model version; it would only apply to a separately configured Model Serving endpoint. - C & D use the fabricated token
stage-production, which is not a recognized URI format in MLflow or Databricks. - E is the most tempting distractor - the path
/model/recommender/Production/invocationslooks plausible, but the Databricks REST API does not accept a stage alias where a version integer is expected.
Memory tip: Think of it this way - the invocation URL slot is typed as an integer (version number), not a string. Stage names like "Production" are metadata labels in the Registry, not routable path components. If a question gives you a stage but no version, the answer involves finding the version first.
Community Discussion
No community discussion yet for this question.