swPredictor: A data-driven performance model for distributed data parallelism training on large-scale HPC clusters

分类与摘要

关注大规模并行算法、超算平台和性能优化，是 HPC 方向的重要条目。

证据摘录：llelism training on large-scale HPC clusters Xianyu Zhu a , Ruohan Wu a , Junshi Chen a,b ,∗, Hong An a,b a University of Science and Technology of China, Hefei, Anhui, China b Laoshan Laboratory, Qingdao, China A R T I C L E I N F O Keywords: High-performance computing Performance modeling Deep learning Distributed training A B S T R A C T Given the complexity of heterogeneous architectures and multi-node col || llelism training on large-scale HPC clusters Xianyu Zhu a , Ruohan Wu a , Junshi Chen a,b ,∗, Hong An a,b a University of Science and Technology of China, Hefei, Anhui, China b Laoshan Laboratory, Qingdao, China A R T I C L E I N F O Keywords: High-performance computing Performance modeling Deep learning Distributed training A B S T R A C T Given the complexity of heterogeneous architectures and mul || pported by the Chinese Academy of Sciences (Strategic Priority Research Program, China), Grant (XDB0500102). And this work is financially supported by Laoshan Laboratory (China) (LSKJ202300305). We would like to thank Ziyue You from the University of Edinburgh, Shengtao Xue from the Northwest A&F University and Jiahao Huang from the National University of Singapore, for their valuable feedback and assistanc || pported by the Chinese Academy of Sciences (Strategic Priority Research Program, China), Grant (XDB0500102). And this work is financially supported by Laoshan Laboratory (China) (LSKJ202300305). We would like to thank Ziyue You from the University of Edinburgh, Shengtao Xue from the Northwest A&F University and Jiahao Huang from the National University of Singapore, for their valuable feedback and

引用

Zhu, Xianyu; Wu, Ruohan; Chen, Junshi; An, Hong，swPredictor: A data-driven performance model for distributed data parallelism training on large-scale HPC clusters，PERFORMANCE EVALUATION，VOL170，November ,2025 （CCF B）

@article{acsa2025_4,
  title = {swPredictor: A data-driven performance model for distributed data parallelism training on large-scale HPC clusters},
  year = {2025},
  doi = {10.1016/j.peva.2025.102530}
}