%0 Conference Paper
%A John, Chelsea Maria
%A Herten, Andreas
%A Penke, Carolin
%A Nassyr, Stepan
%T Performance and Power: Systematic Evaluation of AI Workloads on Accelerators with CARAML
%M FZJ-2024-06447
%P Nan
%D 2024
%Z Also available at: https://arxiv.org/abs/2409.12994
%X The rapid advancement of machine learning (ML) technologies has driven the development of specialized hardware accelerators designed to facilitate more efficient model training. This paper introduces the CARAML benchmark suite, which is employed to assess performance and energy consumption during the training of transformer-based large language models and computer vision models on a range of hardware accelerators, including systems from NVIDIA, AMD, and Graphcore. CARAML provides a compact, automated, extensible, and reproducible framework for assessing the performance and energy of ML workloads across various novel hardware architectures. The design and implementation of CARAML, along with a custom power measurement tool called jpwr, are discussed in detail.
%B Supercomputing Conference 2024, 2024 International Workshop on Performance, Portability, and Productivity in HPC
%C 17 Nov 2024 - 22 Nov 2024, Atlanta (USA)
Y2 17 Nov 2024 - 22 Nov 2024
M2 Atlanta, USA
%F PUB:(DE-HGF)8
%9 Contribution to a conference proceedings
%U <Go to ISI:>//WOS:001451792300127
%R 10.1109/SCW63240.2024.00158
%U https://juser.fz-juelich.de/record/1033567