‹ All AI-300 questions

AI-300 Practice Question 95: Sweep Max Concurrent Trials

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 10 · question 5 of 10

Implement machine learning model lifecycle and operationsmedium

A sweep job is configured for 40 trials on a cluster that the team wants to limit to 4 trials running at a time, to stay within quota. Each trial runs on one node. Which sweep limit setting achieves this?

Select one answer
Show answer and explanation

Answer:

  • A. Set max_concurrent_trials to 4

In a sweep job, max_total_trials sets how many trials run in all, and max_concurrent_trials sets how many may run at once. When it is not specified, all trials are launched in parallel, gated only by compute resources. Setting it to 4 keeps 40 trials to 4 at a time (option A).

- B is wrong because this would stop the sweep after only 4 trials in total, instead of 40 trials 4 at a time. - C is wrong because this delays the evaluation of the early termination policy. It does not limit how many trials run in parallel. - D is wrong because this would run each trial across 4 nodes, using more resources rather than fewer.

Reference: Hyperparameter tuning a model

Practice Test 10 ›

Study guides for AI-300