AWS Certified Machine Learning Engineer - Associate MLA-C01 — Question 72

An ML engineer is using Amazon SageMaker to train a deep learning model that requires distributed training. After some training attempts, the ML engineer observes that the instances are not performing as expected. The ML engineer identifies communication overhead between the training instances. What should the ML engineer do to MINIMIZE the communication overhead between the instances?