"Vocabulary: MLOps and Evaluation"

1. What is the purpose of a model registry in MLOps?

To store raw training data To keep a versioned inventory of models To evaluate model performance To deploy models in production

2. What is evaluated as part of the evaluation (eval) process?

Initial training data Known-correct outputs against curated inputs Model deployment strategies User engagement metrics

3. Which of the following best describes drift in the context of deployed models?

Prompts that change over time The world changing under a deployed model The initial training data losing relevance The process of re-indexing a model

4. What does shadow deployment involve?

Deploying without monitoring Measuring agreement with existing processes without showing outputs Deploying a model to all users at once Storing all model versions for future reference

Continue: "Vocabulary: Classical Machine Learning" → Review this lesson Retake quiz