arrow
返回

ALBERTA: ALgorithm-Based Error Resilience in Transformer Architectures

delete2025-01-01
delete1
delete
OA
AI
H
H. Liu *
V
Vasu Singh
M
Michał Filipiuk
S
Siva Kumar Sastry Hari
DOI:10.1109/OJCS.2024.3400696delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
Vision Transformers are being increasingly deployed in safety-critical applications that demand high reliability. Ensuring the correct execution of these models in GPUs is critical, despite the potential for transient hardware errors. We propose a novel algorithm-based resilience framework called ALBERTA that allows us to perform end-to-end resilience analysis and protection of transformer-based architectures. First, our work develops an efficient process of computing and ranking the resilience of transformers layers. Due to the large size of transformer models, applying traditional network redundancy to a subset of the most vulnerable layers provides high error coverage albeit with impractically high overhead. We address this shortcoming by providing a software-directed, checksum-based error detection technique aimed at protecting the most vulnerable general matrix multiply (GEMM) layers in the transformer models that use either floating-point or integer arithmetic. Results show that our approach achieves over 99% coverage for errors (single bit-flip fault model) that result in a mismatch with < 0.2% and < 0.01% computation and memory overheads, respectively. Lastly, we present the applicability of our framework in various modern GPU architectures under different numerical precisions. We introduce an efficient self-correction mechanism for resolving erroneous detection with an average of less than 2% overhead per error.
Keyword:
Transformers
Resilience
Computational modeling
Hardware
Computer architecture
Transient analysis
Measurement
Correction
detection
protection
resilience
robustness
safety
transformers

期刊

I
IEEE Open Journal of the Computer Society
IF:
8.2
论文数:
411
被引数:
810

机构

N
nvidia corporation
学者数:
767
论文数: 439
被引数: 1
引用论文

引用论文

Molecularly Imprinted Polymer‐Based Sensors for Amperometric Determination of Nonelectroactive Substances
err2009-02-12
err0
PREAI
errMaria Pesavento; Girolamo D'Agostino; Raffaela Biesuz; Giancarla Alberti
err分享
err收藏
Transformers in Vision: A Survey视觉中的变形金刚: 一项调查
err2022-09-13
err1.1K
errOAAI
errKhan, Salman; Naseer, Muzammal; Hayat, Munawar; Zamir, Syed Waqas; Khan, Fahad Shahbaz; Shah, Mubarak
err分享
err收藏
Clinical relevance of cytoskeleton associated proteins for ovarian cancer
err2018-08-09
err0
PREAI
errJohanna Schiewek; Udo Schumacher; Tobias Lange; Simon A. Joosse; Harriet Wikman; Klaus Pantel; Marina Mikhaylova; Matthias Kneussel; Stefan Linder; Barbara Schmalfeldt; Leticia Oliveira-Ferrer; Sabine Windhorst
err分享
err收藏
学者 查看更多内容