一句话：摘要未说明具体贡献，仅指出 GPU 全局显存稀缺限制大模型训练与 DNN 架构探索。
近年来深度学习在多个领域取得巨大成功，其关键之一是更大、更深的深度神经网络（DNN）带来了很高精度。然而，GPU 全局内存是稀缺资源，大模型在训练过程中对内存的需求也随之显著增加，成为重要挑战。这种显存限制削弱了 DNN 架构探索的灵活性。所给摘要仅描述了上述问题背景，未说明 Capuchin 的具体方法、系统设计、实现细节或评估结果，因此无法从摘要确定其技术贡献与实验结论。
In recent years, deep learning has gained unprecedented success in various domains, the key of the success is the larger and deeper deep neural networks (DNNs) that achieved very high accuracy. On the other side, since GPU global memory is a scarce resource, large models also pose a significant challenge due to memory requirement in the training process. This restriction limits the DNN architecture exploration flexibility.
DOI 原文 ·
@inproceedings{paperbot5478,
title = {Capuchin},
author = {Xuan Peng and Xuanhua Shi and Hulin Dai and Hai Jin and Weiliang Ma and Qian Xiong and Fan Yang and Xuehai Qian},
booktitle = {ASPLOS},
year = {2020},
doi = {10.1145/3373376.3378505}
}