arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.22136cs.LGcs.AR
收藏

SigmaQuant: 面向边缘 DNN 推理的硬件感知异构量化方法

SigmaQuant: Hardware-Aware Heterogeneous Quantization Method for Edge DNN Inference

Embedded Systems Laboratory (ESL), EPFL, Switzerland(嵌入式系统实验室(ESL),EPFL,瑞士);University of Applied Sciences and Arts Western Switzerland (HES-SO), Switzerland(应用科学与艺术西瑞士大学(HES-SO),瑞士)

Qunyou Liu, Pengbo Yu, Marina Zapater, David Atienza

AI总结:

SigmaQuant 是一种面向边缘 DNN 推理的自适应异构量化方法,通过动态分配不同比特宽度以优化精度与资源使用平衡。

摘要:

深度神经网络(DNNs)对于在边缘或移动设备上执行高级任务至关重要,但其部署常受到严重资源限制的阻碍,包括有限的内存、能量和计算能力。虽然均匀量化提供了一种压缩模型和减少硬件需求的简便方法,但其未能充分利用各层之间的差异鲁棒性,且在低比特宽度下常导致精度下降或资源使用不优。相比之下,异构量化通过为各层分配不同比特宽度可缓解这些缺点。然而,当前异构量化方法要么需要庞大的设计空间搜索,要么缺乏对不同硬件条件(如内存大小、能量预算和延迟要求)的适应性。为填补这些空白,本文提出 SigmaQuant,一种自适应的层间异构量化框架,旨在高效平衡精度和资源使用,以适应不同的边缘环境,而无需进行穷举搜索。

英文摘要:

Deep neural networks (DNNs) are essential for performing advanced tasks on edge or mobile devices, yet their deployment is often hindered by severe resource constraints, including limited memory, energy, and computational power. While uniform quantization provides a straightforward approach to compress model and reduce hardware requirement, it fails to fully leverage the varying robustness across layers, and often lead to accuracy degradation or suboptimal resource usage, particularly at low bitwidths. In contrast, heterogeneous quantization, which allocates different bitwidths to individual layers, can mitigate these drawbacks. Nonetheless, current heterogeneous quantization methods either needs huge brute-force design space search or lacks the adaptability to meet different hardware conditions, such as memory size, energy budget, and latency requirement. Filling these gaps, this work introduces \textbf{\textit{SigmaQuant}}, an adaptive layer-wise heterogeneous quantization framework designed to efficiently balance accuracy and resource usage for varied edge environments without exhaustive search.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑