Intel Gaudi Profiler User Guide

Profiling is the process of measuring and analyzing the performance of a software application to understand its resource usage and identify potential bottlenecks and optimization opportunities. It can involve tracking metrics such as CPU and HPU memory usage, execution time of functions, and system resource utilization.

Before you get started, review the below documents highlighting the Profiling workflow and real-world examples.

If you are already using PyTorch with TensorBoard, you can easily start Profiling with PyTorch for both model development and profiling by replacing GPU with HPU in your sources. For low-level focused profiling or writing models in Intel Gaudi software without a framework, use the Profiling with Intel Gaudi software.