AI-300 Practice Question 84: Local Deployment Debug

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 9 · question 4 of 10

Implement machine learning model lifecycle and operationseasy

Every time the team fixes the scoring script of an online deployment, they wait many minutes for a cloud deployment just to see the next error. Which practice gives a faster feedback loop?

Select one answer
Show answer and explanation

Answer:

  • A. Deploy locally with the local option and debug the script there

Azure Machine Learning can create an online endpoint and deployment on the local machine, using Docker, with the --local option in the CLI or local=True in the SDK. It is meant for development and testing, so scoring script errors are found and fixed in seconds before the cloud deployment (option A).

- B is wrong because more instances do not shorten container build and startup, and the error would be the same. - C is wrong because a batch endpoint has a different scoring interface and still runs in the cloud, so the delay remains. - D is wrong because autoscale changes the number of instances under load. It does not speed up finding script errors.

Reference: Deploy and score a machine learning model by using an online endpoint

Practice Test 9 ›

Study guides for AI-300