A Disaggregated Memory System for Deep Learning
A Disaggregated Memory System for Deep Learning
复制标题
用于深度学习的分解内存系统
DOI：
10.1109/mm.2019.2929165
复制
发表时间：
2019
期刊：
影响因子：
3.6
通讯作者：
Minsoo Rhu
中科院分区：
文献类型：
--
作者：
Youngeun Kwon;Minsoo Rhu
随着深度学习的复杂性（DL）模型的扩展，计算机架构师面临着内存“容量”墙，其中加速器设备内部的有限的物理内存会约束可以训练和部署的算法。在设计以加速器为中心的DL内存系统时。
As the complexity of deep learning (DL) models scales up, computer architects are faced with a memory “capacity” wall, where the limited physical memory inside the accelerator device constrains the algorithm that can be trained and deployed. This article summarizes our recent work on designing an accelerator-centric, disaggregated memory system for DL.