Performance Analysis of AI and HPC Workloads

Content

With the increasing adoption of AI technologies, evaluating computational performance of AI applications in HPC systems has become critical for results improvement. In this course we shall use HPC performance tools to profile and evaluate performance bottlenecks in deep neural networks and learn tips for efficient training and deployment of deep learning models in HPC systems. Tools to be covered in this course include: Nvidia Nsight System, Score-P and Vampir.

Requirements

• Practical knowledge of deep learning frameworks, Tensorflow or Pytorch
• Programming skills in Python
• Knowledge in Linux

Learning goal

• The course is intended to equip participants with fundamental knowledge on how to efficiently use HPC systems to run AI applications.

Skills

Trainer

• Hauke Kirchner
• Dr. Jack Ogaja
• Dr. Kevin Lüdemann

Next appointment