A land-records vault hosts twenty rarely used classifiers on one SageMaker multi-model endpoint. Staff want one Model Monitor schedule that reports drift for each classifier. What should they do?
Select an answer to reveal the explanation.
Short Explanation
Twenty classifiers share one multi-model endpoint. Model Monitor will not give a per-model drift report on that shared host. Put each model on a single-model or pipeline endpoint if it needs its own monitor.
Full Explanation
SageMaker Model Monitor supports single-model endpoints and can watch an inference pipeline as a whole. It does not provide per-model monitoring on a multi-model endpoint. Data capture on that shared host does not split statistics by target model, and Rekognition is not a substitute monitor. If each classifier needs its own drift report, host it on a single-model (or pipeline) endpoint first.