ai-300 Practice Question 2016
Exam: ai-300
Domain: Implement CI/CD for machine learning
Topic: Automate model training and deployment
Difficulty: medium
During the deployment of an AI model, you decide to use canary deployment to minimize risk. Which of the following best describes the canary deployment process in the context of Azure Machine Learning and GitHub Actions?
Answer Options
A
Deploy the entire model to all users immediately after training.
B
Deploy a small portion of the traffic to the new model version to test its performance before fully rolling out.
C
Manually test the model on a subset of data before deploying it to production.
D
Deploy the model to a separate environment for testing but not to production.
Correct Answer
B: Deploy a small portion of the traffic to the new model version to test its performance before fully rolling out.
Explanation
The correct answer is B because it accurately describes the canary deployment process where a small percentage of traffic is directed to the new model version to test its performance before full deployment. Option A is incorrect as it suggests deploying the entire model at once, which contradicts the purpose of canary deployment. Option C is incorrect as it implies that canary deployment is not automated, which is not true. Option D is incorrect as it suggests that canary deployment is not used for testing, which is a key aspect of canary deployment.