arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.22352cs.ARcs.AI
收藏

GRAU:面向神经网络硬件加速器的通用可重构激活单元设计

GRAU: A Generic Runtime-Reconfigurable Activation-Requantization Unit for Quantized Neural Network Accelerators

Ruhr University Bochum, Germany(博德姆鲁尔大学);Dresden University of Technology, Germany(德累斯顿技术大学);Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI Dresden/Leipzig), Germany(可扩展数据与人工智能中心(ScaDS.AI 德累斯顿/莱比锡))

Yuhao Liu, Salim Ullah, Akash Kumar

AI总结:

提出基于分段线性拟合的可重构激活硬件GRAU,用2的幂近似斜率,仅需比较器和1位移位器,支持混合精度量化和SiLU等非线性函数,相比多阈值激活器减少90%以上LUT消耗,在6-8段时达到最佳权衡。

摘要:

随着神经网络规模的持续增长,低精度量化被广泛用于边缘加速器。经典的多阈值激活硬件需要2^n个阈值来实现$n$位输出,导致硬件成本随精度提高而迅速增加。我们提出了一种基于分段线性拟合的可重构激活硬件GRAU,其中分段斜率由2的幂近似。我们的设计仅需要基本的比较器和1位右移器,支持混合精度量化和SiLU等非线性函数。与多阈值激活器相比,GRAU减少了90%以上的LUT消耗,实现了更高的硬件效率、灵活性和可扩展性。最佳权衡通常在6-8段时实现,而在激进的低成本设置下,复杂的非线性可能会遭受更大的精度下降。

英文摘要:

As neural networks continue to scale, low-precision quantization is widely adopted in edge accelerators, but conventional multi-threshold activation hardware requires 2^n-1 thresholds for n-bit integer outputs, leading to rapidly increasing hardware costs. We propose GRAU, a Generic Runtime-Reconfigurable Activation-Requantization Unit that fuses batch normalization, nonlinear activation, and requantization through piecewise-linear fitting with Power-of-Two slopes. GRAU requires only basic comparators and 1-bit right shifters, while supporting mixed-precision quantization and runtime switching among post-accumulation functions, including modern nonlinear activations such as SiLU and GELU. Compared with multi-threshold activation-requantization units, GRAU reduces LUT consumption by approximately 90% or more while improving hardware efficiency, flexibility, and scalability. We further explore the trade-off between accuracy and hardware cost across different approximation settings and provide an open-source toolchain for GRAU-aware network training.

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。

↑