A hands-on workshop on how to harness the power of high-performance computing (HPC) clusters for efficient deep learning training. This workshop is designed for PhD students with an interest in leveraging HPC resources to accelerate their machine learning research.
Description: The workshop will explore the intricacies of harnessing High-Performance Computing (HPC) systems for machine learning applications, focusing on the adept utilization of SLURM-managed clusters. Participants will delve into the nuances of efficiently training deep neural models using multiple GPUs and compute nodes. Through a blend of theoretical foundations and practical demonstrations, attendees will acquire the skills needed to optimize machine learning model training on HPC clusters. This session is tailored to offer insights for researchers and PhD students seeking to elevate their machine learning capabilities within an HPC framework.
Workflow: The workshop is scheduled for a virtual session via Zoom. Attendees will gain access to the supercomputer cluster through SSH.
Difficulty: Intermediate
Language: English
Recommended prior knowledge: Basics of machine learning and Python
Target audience: PhD students
Knowledge gained from training:
• Gain an understanding of HPC architecture and how to utilize SLURM to manage HPC resources for deep learning tasks.
• Discover techniques for optimizing deep neural model training on multi-GPU and multi-node HPC clusters.
• Apply theoretical concepts to practical examples and hands-on exercises to enhance your deep learning skills.
Location:
1.) Zoom: https://uni-lj-si.zoom.us/j/94656999191?pwd=K0o0UkNJS3pVL1F6WEJMM1Y5MFhGUT09
Organiser:
University of Ljubljana, Faculty of Electrical Engineering
Lecturers:
Janez Perš
Janez Perš is an Associate Professor at the Faculty of Electrical Engineering at the University of Ljubljana. His research areas are computer, machine, and robotic vision, parallel and distributed systems, and human movement analysis. He is the course holder for Embedded Systems, Computer Vision, Imaging Technologies, and Communications in Automation.
E-mail: janez.pers@fe.uni-lj.si
Janez Križaj
Janez Križaj is a researcher at the Faculty of Electrical Engineering at the University of Ljubljana. His research areas are deep learning, computer vision, biometrics, face recognition, pattern recognition, and image processing.
E-mail: janez.krizaj@fe.uni-lj.si
Twitter
