CERTIFIED-MACHINE-LEARNING-PROFESSIONAL Question #18: Real Exam Question with Answer & Explanation

Question

Which of the following describes the concept of MLflow Model flavors?

Options

• AA convention that deployment tools can use to wrap preprocessing logic into a Model
• BA convention that MLflow Model Registry can use to version models
• CA convention that MLflow Experiments can use to organize their Runs by project
• DA convention that deployment tools can use to understand the model
• EA convention that MLflow Model Registry can use to organize its Models by project

Explanation

Warning: The marked correct answer appears to be wrong. Option C describes how MLflow Experiments organize Runs - that has nothing to do with Model flavors. The actual correct answer is D.

MLflow Model flavors are a standardized convention that tells deployment tools how to interpret and load a model. Each saved MLflow model can expose one or more flavors (e.g., python_function, sklearn, tensorflow), and deployment infrastructure reads these to know which runtime and loading method to use - enabling framework-agnostic serving.

Why the distractors are wrong:

• A - Flavors don't wrap preprocessing logic; that's closer to what a Pipeline or custom pyfunc model does.
• B - Model Registry versioning is a separate concept; flavors are about how a model is consumed, not how versions are tracked.
• C - Organizing Runs by project is the job of Experiments and their naming/tagging system - completely unrelated to flavors.
• E - Model Registry organizes models by name/version/stage, not by project via flavors.

Memory tip: Think of a model as a dish that can come in multiple flavors (sklearn, TF, PyTorch). The waiter (deployment tool) reads the available flavors off the menu to know how to serve it. Flavors = deployment interface contract.

Double-check your source material - if this appeared on an official practice exam, it may contain an error.

No community discussion yet for this question.