Model Deployment & Monitoring Quiz - Competitive Exams GCP_ML_Engineer

Free practice quiz: Model Deployment & Monitoring (Competitive Exams GCP_ML_Engineer). 61 multiple-choice questions with answers and explanations. Play online, no sign-up needed.

This Competitive Exams GCP_ML_Engineer practice quiz has 61 multiple-choice questions on Model Deployment & Monitoring. Try the 5 sample questions below (tap an option to check yourself), then play the full quiz online. No sign-up needed.

Sample questions

1. Which Vertex AI resource is used to deploy a trained model for online predictions with autoscaling capabilities?

   1. Model Registry
   2. Endpoint
   3. Pipeline
   4. Feature Store
   Show answer

   Answer: B. Endpoint

2. Which metric does Vertex AI Model Monitoring use to detect prediction drift when ground truth labels are unavailable?

   1. Jensen-Shannon divergence on feature distributions
   2. Root mean squared error
   3. Mean absolute error
   4. Precision-Recall AUC
   Show answer

   Answer: A. Jensen-Shannon divergence on feature distributions

3. When using Vertex AI Model Monitoring, what is the purpose of the monitoring objective?

   1. To indicate which metric to track (e.g., feature skew, prediction drift)
   2. To set the number of monitoring replicas
   3. To define the frequency of batch predictions
   4. To specify the model architecture to monitor
   Show answer

   Answer: A. To indicate which metric to track (e.g., feature skew, prediction drift)

4. In Vertex AI, what is the purpose of the 'explanation config' on a deployed model?

   1. To define the model's input schema
   2. To enable feature attributions for each prediction
   3. To set up autoscaling
   4. To log all prediction requests
   Show answer

   Answer: B. To enable feature attributions for each prediction

5. Which Vertex AI feature allows you to organize multiple models and their versions for deployment and monitoring?

   1. Pipeline
   2. Experiment
   3. Endpoint
   4. Model Registry
   Show answer

   Answer: D. Model Registry

More questions in this quiz

Answers and explanations are in the full quiz.

1. What is the primary purpose of a Vertex AI batch prediction job?
2. In Vertex AI, which setting defines the container image and command used to serve predictions on an endpoint?
3. Which Vertex AI feature automatically detects anomalies in incoming prediction requests?
4. What does Vertex AI Model Monitoring compare to detect training-serving skew?
5. Which component must be created before you can log model monitoring metrics to Cloud Storage in Vertex AI?
6. When deploying a model to a Vertex AI endpoint, what does traffic split enable?
7. Which technique allows deploying a new model version to an endpoint while limiting exposure to a small fraction of users?
8. What is the key characteristic of a shadow deployment on Vertex AI?
9. Which Vertex AI setting controls the vertical scaling of resources allocated to a deployed model?
10. What does setting minReplicaCount to 1 on a Vertex AI deployed model ensure?
11. Which artifact is required to enable Vertex AI Model Monitoring for a deployed model?
12. What is the effect of configuring an endpoint with a dedicated private IP (Private Service Connect) instead of a public endpoint?
13. Which Vertex AI feature allows you to test a model's predictions interactively before deploying it to an endpoint?
14. Which IAM permission is required to deploy a model to a Vertex AI endpoint?
15. Which Vertex AI resource stores a model's lineage, version, and associated metadata for governance?
16. Which method is used to send online prediction requests to a Vertex AI endpoint?
17. Which type of container is typically used for serving a custom model on Vertex AI?
18. What is the role of the health route in a Vertex AI custom prediction container?
19. Which Vertex AI feature can automatically detect and alert on feature skew between training and serving data?
20. Which deployment strategy involves running the old and new model simultaneously and comparing their predictions on live traffic without exposing users to the new model?
21. Which configuration on a Vertex AI endpoint enables request-response logging to BigQuery for auditing and monitoring?
22. Which Vertex AI component is responsible for defining the schema of the instances sent to a deployed model, enabling request validation and feature extraction?
23. What is the purpose of the Vertex AI endpoint's IAM policy?
24. In Vertex AI, which feature helps you understand which input features contributed most to a prediction for a deployed model?
25. Which Vertex AI resource can be used to organize multiple models and their versions for deployment and lifecycle management?
26. When deploying a model to a Vertex AI endpoint, what does the 'machine type' setting determine?
27. Which Vertex AI endpoint feature allows you to direct a percentage of prediction traffic to different model versions deployed on the same endpoint?
28. What is required in a custom prediction container to handle prediction requests on Vertex AI?
29. Which Vertex AI feature allows you to schedule regular batch predictions and automatically store results in BigQuery?
30. What is the effect of setting a model's deployment resource to use a GPU?
31. What is the purpose of the Vertex AI endpoint's 'model deployment' settings for 'min replica count'?
32. Which of the following is a valid method to send online prediction requests to a Vertex AI endpoint?
33. Which Vertex AI monitoring feature compares the distribution of a feature in production to its distribution in the training dataset?
34. Which Vertex AI resource is used to store a model's version and metadata for governance and lineage tracking?
35. Which deployment strategy involves running a new model alongside the existing one without affecting live traffic, until it is validated?
36. Which Vertex AI endpoint setting enables request-response logging to BigQuery for auditing and analysis?
37. What is the primary purpose of Vertex AI Model Monitoring?
38. Which Vertex AI feature allows you to use a private IP address for your endpoint to restrict access to your VPC network?
39. Which Vertex AI endpoint type is best suited for real-time, low-latency predictions that require a persistent serving infrastructure?
40. Which Vertex AI resource is used to store model versions and their metadata for governance?
41. What is the effect of enabling request-response logging on a Vertex AI endpoint?
42. Which Vertex AI deployment strategy involves running the old and new model simultaneously and comparing their predictions without affecting users?
43. What is the purpose of the 'monitoring objective' in Vertex AI Model Monitoring?
44. Which Vertex AI endpoint setting controls the number of replicas that are always running?
45. Which Vertex AI resource is responsible for capturing the raw input and output tensors of a deployed model for later retrieval?
46. In Vertex AI, which configuration allows the endpoint to return feature attributions along with predictions for models that support explainability?
47. Which type of Vertex AI endpoint is designed to handle high-throughput, low-latency inference requests using a shared, managed infrastructure?
48. What is the purpose of the 'deployed model' object in Vertex AI?
49. Which Vertex AI feature allows you to define a custom preprocessing step that runs before the model's predict function in a custom container?
50. In Vertex AI, what is the effect of setting an endpoint's 'traffic split' to send 100% of traffic to one deployed model?
51. What is the role of the 'instance schema' when configuring a Vertex AI deployed model?
52. When using Vertex AI Model Monitoring, what does the 'monitoring objective' define?
53. Which Vertex AI endpoint setting controls the maximum number of replicas that can be provisioned during autoscaling?
54. In Vertex AI, which deployment strategy involves running a new model alongside the existing one without affecting live traffic, and comparing their predictions?
55. What is the purpose of the 'explanation config' on a deployed model in Vertex AI?
56. Which Vertex AI endpoint feature allows you to direct a percentage of prediction traffic to different model versions?

More quizzes

• Vertex AI & Pipelines▶ Play
• Data Engineering for ML▶ Play
• ML Problem Framing▶ Play

Browse all GCP_ML_Engineer quizzes

More to play