AI-300 Practice Question 87: Developer Tier Evaluation

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 9 · question 7 of 10

Design and implement a GenAIOps infrastructurehard

A team fine-tuned five candidate models and wants reviewers to test each one for a few hours to pick a winner. The candidates will not serve production traffic, and the team wants to avoid an hourly hosting fee while they are compared. Which deployment type should they use?

Select one answer
Show answer and explanation

Answer:

  • D. The Developer deployment type, which has no hourly hosting fee and no availability SLA

Developer deployments of fine-tuned models offer an experience similar to Global Standard, without an hourly hosting fee and without an availability SLA. They are designed for evaluating model candidates, not for production, and each deployment lasts 24 hours, which fits a few hours of review (option D).

- A is wrong because a Standard fine-tuned deployment is billed hourly while it exists, which is the cost the team wants to avoid. - B is wrong because provisioned throughput reserves capacity and is billed for the reservation, which is far more than an evaluation needs. - C is wrong because batch is asynchronous with a target of 24 hours, so reviewers could not interactively test the outputs.

Reference: Deploy a fine-tuned model

Practice Test 9 ›

Study guides for AI-300