EUVIMEDEUVIMED entdecken
Crossref

GQoD: Group-aware quantization for efficient layer-wise pipelined inference of Multi-Exit BERT on end devices

Zonglin Zou, Jinghui Zhang, Wei Zhang, Weilong Xin, Ding Ding, Jiahui Jin

Journal of Systems Architecture · 2026-07 · 176 · 103826

DOI: 10.1016/j.sysarc.2026.103826 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Systems Architecture, 2026-07. Verfasst von Zonglin Zou und 5 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Jiangsu Province Department of Science and Technology, Fundamental Research Funds for the Central Universities.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Kein Abstract verfügbar.

Förderung

  • National Natural Science Foundation of China
  • Jiangsu Province Department of Science and Technology
  • Fundamental Research Funds for the Central Universities
  • Ministry of Education of the People's Republic of China
  • Key Laboratory of Computer Network and Information Integration, Ministry of Education

Zitieren

APA

Zou, Z., Zhang, J., Zhang, W., Xin, W., Ding, D., & Jin, J. (2026). GQoD: Group-aware quantization for efficient layer-wise pipelined inference of Multi-Exit BERT on end devices. Journal of Systems Architecture, 176, 103826. https://doi.org/10.1016/j.sysarc.2026.103826

Vancouver

Zou Z, Zhang J, Zhang W, Xin W, Ding D, Jin J. GQoD: Group-aware quantization for efficient layer-wise pipelined inference of Multi-Exit BERT on end devices. Journal of Systems Architecture. 2026;176:103826. doi:10.1016/j.sysarc.2026.103826

BibTeX
@article{Zou2026,
  author = {Zou, Zonglin and Zhang, Jinghui and Zhang, Wei and Xin, Weilong and Ding, Ding and Jin, Jiahui},
  title = {GQoD: Group-aware quantization for efficient layer-wise pipelined inference of Multi-Exit BERT on end devices},
  journal = {Journal of Systems Architecture},
  year = {2026},
  volume = {176},
  pages = {103826},
  doi = {10.1016/j.sysarc.2026.103826}
}
Kostenlos starten →