← 返回 JSSC 论文列表JSSC 2019第1期Digital Circuits65nmNeural Network Accelerator
UNPU- An Energy-Efficient Deep Neural Network Accelerator With Fully Variable We
提出一种能效优化的深度神经网络加速器UNPU，支持可变权重精度和多种网络层类型。
65nm CMOS, 0.63-1.1V, 200MHz, 峰值性能345.6GOPS(16bit)-7372GOPS(1bit)
神经网络加速器能效优化可变精度移动深度学习硬件架构
▸支持1至16位可变权重精度
▸基于查找表的位串行处理单元降低能耗
▸统一架构提升卷积层峰值性能1.15倍
Abstract
An energy-efficient deep neural network (DNN)
accelerator, unified neural processing unit (UNPU), is proposed
for mobile deep learning applications. The UNPU can support
both convolutional layers (CLs) and recurrent or fully connected
layers (FCLs) to support versatile workload combinations to
accelerate various mobile deep learning applications. In addition,
the UNPU is the first DNN accelerator ASIC that can support
fully variable weight bit precision from 1 to 16 bit. It enables
the UNPU to oper