BriefGPT.xyz
Jun, 2023
深度学习硬件准确性恢复的敏感性感知微调
Sensitivity-Aware Finetuning for Accuracy Recovery on Deep Learning Hardware
HTML
PDF
Lakshmi Nair, Darius Bunandar
TL;DR
本文介绍了Sensitivity-Aware Finetuning (SAFT)方法以提高模型精度，并使用该方法进行噪声注入训练，以冻结特定层来提高噪声敏感层的性能，结果表明SAFT可与噪声注入训练方法相媲美，并且速度快2倍至8倍。
Abstract
Existing methods to recover
model accuracy
on
analog-digital hardware
in the presence of quantization and analog noise include
noise-injection tr
→