
Amazon SageMaker Profiler is a new observability tool in preview that provides GPU/CPU profiling insights to help optimize deep learning model training performance and identify hardware utilization bottlenecks.


<div><p>This article announces the preview of Amazon SageMaker Profiler, an observability tool for optimizing deep learning model training performance.</p><ul><li>Provides granular GPU/CPU profiling insights for large deep learning workloads</li><li>Identifies optimization opportunities through GPU/CPU utilization metrics and trace plots</li><li>Detects bottlenecks from uneven resource utilization across instances</li><li>Reduces profiling overhead while supporting longer durations and more instances</li><li>Available in US East (Ohio, N. Virginia), US West (Oregon), and Europe (Frankfurt, Ireland)</li><li>Free during preview period in supported regions</li></ul><p>SageMaker Profiler helps data scientists optimize hardware performance for large-scale distributed training workloads by providing detailed performance visibility.</p></div>


Related articles