FAST: DNN Training Under Variable Precision Block Floating Point with Stochastic Rounding
Author:
Affiliation:
1. Harvard University
2. Franklin and Marshall College
Funder
Air Force Research Laboratory
Defense Advanced Research Projects Agency
Publisher
IEEE
Link
http://xplorestaging.ieee.org/ielx7/9773179/9773180/09773221.pdf?arnumber=9773221
Reference56 articles.
1. A domain-specific supercomputer for training deep neural networks
2. Adding gradient noise improves learning for very deep networks;neelakantan,2015
3. Procrustes: a Dataflow and Accelerator for Sparse Deep Neural Network Training
4. Eager pruning
5. TensorDash: Exploiting Sparsity to Accelerate Deep Neural Network Training
Cited by 24 articles. 订阅此论文施引文献 订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献
1. Optimization of block-scaled integer GeMMs for efficient DNN deployment on scalable in-order vector processors;Journal of Systems Architecture;2024-09
2. A Dynamic Execution Neural Network Processor for Fine-Grained Mixed-Precision Model Training Based on Online Quantization Sensitivity Analysis;IEEE Journal of Solid-State Circuits;2024-09
3. ABS: Accumulation Bit-Width Scaling Method for Designing Low-Precision Tensor Core;IEEE Transactions on Very Large Scale Integration (VLSI) Systems;2024-09
4. Contemporary Advances in Neural Network Quantization: A Survey;2024 International Joint Conference on Neural Networks (IJCNN);2024-06-30
5. DACAPO: Accelerating Continuous Learning in Autonomous Systems for Video Analytics;2024 ACM/IEEE 51st Annual International Symposium on Computer Architecture (ISCA);2024-06-29
1.学者识别学者识别
2.学术分析学术分析
3.人才评估人才评估
"同舟云学术"是以全球学者为主线，采集、加工和组织学术论文而形成的新型学术文献查询和分析系统，可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容，当前同舟云学术共收录了国内外主流学术期刊6万余种，收集的期刊论文及会议论文总量共计约1.5亿篇，并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询！咨询电话：010-8811{复制后删除}0370
www.globalauthorid.com
TOP
Copyright © 2019-2024 北京同舟云网络信息技术有限公司 京公网安备11010802033243号 京ICP备18003416号-3