Toward Low-Bit Neural Network Training Accelerator by Dynamic Group Accumulation

Yixiong Yang,Ruoyang Liu,Wenyu Sun,Jinshan Yue,Huazhong Yang,Yongpan Liu

2022 27th Asia and South Pacific Design Automation Conference (ASP-DAC)（2022）

引用 0|浏览5

暂无评分

摘要

Low-bit quantization is a big challenge for neural network training. Conventional training hardware adopts FP32 to accumulate the partial-sum result, which seriously degrades energy efficiency. In this paper, a technology called dynamic group accumulation (DGA) is proposed to reduce the accumulation error. First, we model the proposed group accumulation method and give the optimal DGA algorithm. S...

查看译文

关键词

Training,Power demand,Quantization (signal),Design automation,Heuristic algorithms,Neural networks,Asia

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要