arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2306.09481cs.ARcs.ETcs.LGcs.NE
收藏

利用剩余数系设计高精度模拟深度神经网络加速器

Leveraging Residue Number System for Designing High-Precision Analog Deep Neural Network Accelerators

Boston University(波士顿大学);Harvard University(哈佛大学);Lightmatter

Cansu Demirkiran, Rashmi Agrawal, Vijay Janapa Reddi, Darius Bunandar, Ajay Joshi

AI总结:

针对模拟DNN加速器中高精度数据转换器成本高的问题,提出利用剩余数系(RNS)组合低精度运算实现高精度,还采用冗余RNS实现容错,可大幅降低数据转换器能耗。

摘要:

在模拟DNN(深度神经网络)加速器中,实现高精度的同时保持良好能效颇具挑战,因为高精度数据转换器成本高昂。本文通过使用剩余数系(RNS),将多个低精度运算组合为高精度运算,以此攻克这一难题,消除了由ADC(模数转换器)精度有限导致的信息损失。研究表明,仅使用6位精度的数据转换器,RNS就能让前沿DNN推理达到99%的FP32精度。我们提出采用冗余RNS来实现具备容错能力的模拟加速器。此外,研究显示,与常规定点方法相比,RNS可将模拟加速器内数据转换器的能耗降低数个数量级。

英文摘要:

Achieving high accuracy, while maintaining good energy efficiency, in analog DNN accelerators is challenging as high-precision data converters are expensive. In this paper, we overcome this challenge by using the residue number system (RNS) to compose high-precision operations from multiple low-precision operations. This enables us to eliminate the information loss caused by the limited precision of the ADCs. Our study shows that RNS can achieve 99% FP32 accuracy for state-of-the-art DNN inference using data converters with only $6$-bit precision. We propose using redundant RNS to achieve a fault-tolerant analog accelerator. In addition, we show that RNS can reduce the energy consumption of the data converters within an analog accelerator by several orders of magnitude compared to a regular fixed-point approach.

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。

↑