H3D-Transformer: A Heterogeneous 3D (H3D) Computing Platform for Transformer Model Acceleration on Edge Devices
ACM Transactions on Design Automation of Electronic Systems · 31.05.2024 · 29 (3) · 1-19
DOI: 10.1145/3649219 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Design Automation of Electronic Systems, online erschienen am 22.04.2024. Verfasst von Yandong Luo und 1 weiteren Autor:innen. Die Arbeit wurde bisher 34-mal zitiert (Crossref) und nennt 45 Literaturangaben. Gefördert von: CHIMES, SRC/DARPA JUMP 2.0 centers.
Abstract (Originalsprache)
Prior hardware accelerator designs primarily focused on single-chip solutions for 10 MB-class computer vision models. The GB-class transformer models for natural language processing (NLP) impose challenges on existing accelerator design due to the massive number of parameters and the diverse matrix multiplication (MatMul) workloads involved. This work proposes a heterogeneous 3D-based accelerator design for transformer models, which adopts an interposer substrate with multiple 3D memory/logic hybrid cubes optimized for accelerating different MatMul workloads. An approximate computing scheme is proposed to take advantage of heterogeneous computing paradigms of mixed-signal compute-in-memory (CIM) and digital tensor processing units (TPU). From the system-level evaluation results, 10 TOPS/W energy efficiency is achieved for the BERT and GPT2 model, which is about 2.6× ∼ 3.1× higher than the baseline with 7 nm TPU and stacked FeFET memory.
Förderung
- CHIMES
- SRC/DARPA JUMP 2.0 centers
Zitieren
APA
Luo, Y., & Yu, S. (2024). H3D-Transformer: A Heterogeneous 3D (H3D) Computing Platform for Transformer Model Acceleration on Edge Devices. ACM Transactions on Design Automation of Electronic Systems, 29(3), 1-19. https://doi.org/10.1145/3649219
Vancouver
Luo Y, Yu S. H3D-Transformer: A Heterogeneous 3D (H3D) Computing Platform for Transformer Model Acceleration on Edge Devices. ACM Transactions on Design Automation of Electronic Systems. 2024;29(3):1-19. doi:10.1145/3649219
BibTeX
@article{Luo2024,
author = {Luo, Yandong and Yu, Shimeng},
title = {H3D-Transformer: A Heterogeneous 3D (H3D) Computing Platform for Transformer Model Acceleration on Edge Devices},
journal = {ACM Transactions on Design Automation of Electronic Systems},
year = {2024},
volume = {29},
number = {3},
pages = {1--19},
doi = {10.1145/3649219}
}
EUVIMED