Knowledge check

Knowledge check

Complete

Which of the following best explains the benefit of using a Docker container to package an ML model?

A) It increases the model's accuracy during inference.B) It reduces the training time for large datasets.C) It allows models to be packaged with their environment and dependencies for consistent execution.D) It eliminates the need for version control of code and data.

Which Dockerfile instruction defines the command that runs when a container is started?

A) RUNB) WORKDIRC) CMDD) COPY

Which deployment strategy enables safe testing of a new model version by routing a small portion of traffic to it?

A) Blue/green deploymentB) Canary releaseC) Immediate full releaseD) Shadow deployment

What is a key benefit of maintaining a model registry in ML deployment workflows?

A) It stores trained models along with metadata for traceability and reproducibility.B) It visualises feature importance for trained models.C) It replaces the need for Git-based version control.

Which of the following is an example of a versioning strategy for ML models?

A) Saving model files in random folders on a shared driveB) Naming each model file "model_final.pkl" regardless of changesC) Using tools like MLflow or DVC to track model versions, training data, and metricsD) Relying on memory to track what changes were made

Why is it important to monitor data drift in deployed ML systems?

A) To improve training speed on future modelsB) To detect changes in the data that may degrade model performance over timeC) To increase the number of model predictions per secondD.) To test model interpretability

Which of the following would be most useful in a logging strategy for deployed ML models?

A) Storing predictions in unstructured text files for occasional reviewB) Logging confidence scores, inputs, and timestamps in a structured format such as JSONC) Logging only errors without any contextD) Ignoring logs unless the system fails

What is the role of a feedback loop in model monitoring?

A) It collects user complaints about slow load times.B) It triggers container rebuilds after each prediction.C) It collects real-world outcomes and uses them to improve model performance.D) It turns off the model if performance drops.

Which of the following best supports auditability of an ML model used in a regulated industry?

A) A folder containing the most recent model file with no training logsB) A registry entry with dataset version, code commit ID, hyperparameters, and evaluation metricsC) A written summary of team discussions about model performanceD) Weekly model predictions stored without inputs or outcomes

Which documentation format is commonly used to record an ML model's purpose, dataset, limitations, and fairness considerations?

A) CI/CD pipeline configB) API reference guideC) Model cardD) Log file

Evaluate your understanding of this unit by completing the Knowledge check.

Knowledge check

Action item: Knowledge check

In progress