Find the Bottleneck: Optimize AI Pipelines With Nsight Systems

Learn to use NVIDIA Nsight Systems to detect bottlenecks in AI pipelines and explore new features to optimize and scale the execution of your applications. We'll guide you through the performance analysis process of GPU-accelerated AI applications that run across multiple, possibly containerized, compute instances. In addition to profiling of CPU, GPU, network and I/O activity, we'll show how NVTX and the plugin mechanism enable the collection of custom data, which allows an even more comprehensive and tailored analysis on the target platform. You'll also learn about the recipe system that simplifies the identification of bottlenecks in applications running across multiple nodes.
Prerequisite(s):

Familiarity with GPUs and CUDA.