BriefGPT.xyz
Jun, 2024
多项式LUT-Add:基于FPGA的宽输入LUT推理
PolyLUT-Add: FPGA-based LUT Inference with Wide Inputs
HTML
PDF
Binglei Lou, Richard Rademacher, David Boland, Philip H. W. Leong
TL;DR
通过引入PolyLUT-Add技术,结合多个PolyLUT子神经元进行相加以提高准确性,并描述了一个提高可扩展性的新型架构,实验结果表明,在相似准确度的情况下,PolyLUT-Add相比传统方法能够减少1.3-7.7倍的LUT使用量并降低1.2-2.2倍的延迟。
Abstract
fpgas
have distinct advantages as a technology for deploying
deep neural networks
(DNNs) at the edge.
lookup table
(LUT) based networks, w
→