QuestionQ82

VerifiedDeployment and Orchestration of ML Workflows

An ML engineer is running a hyperparameter tuning job in Amazon SageMaker AI. The hyperparameter tuning job will train an XGBoost model using a tabular dataset with millions of rows.

Each individual training job is configured to use up to 8 GPU instances, resulting in high costs if too many jobs run concurrently. The ML engineer needs to ensure that no more than five training jobs run in parallel at any time while still exploring all hyperparameter configurations.

Which solution meets these requirements?

Explanation

In Amazon SageMaker AI hyperparameter tuning, MaxParallelTrainingJobs sets the maximum number of concurrent training jobs that the tuning job can launch. Setting it to 5 limits GPU-intensive jobs to five at once without setting a cap on the total number of training jobs or hyperparameter configurations explored.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!