A BF16 FMA is All You Need for DNN Training-Reference-Cited by-同舟云学术

A BF16 FMA is All You Need for DNN Training

Author:

Osorio John¹^ORCID,Armejach Adria¹^ORCID,Petit Eric²,Henry Greg²,Casas Marc¹

Affiliation:

1. Department of Computer Science, Barcelona Supercomputing Center, Barcelona, Spain

2. Intel Corporation, Portland, OR, USA

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Subject

Computer Science Applications,Human-Computer Interaction,Information Systems,Computer Science (miscellaneous)

Link

Reference40 articles.

1. Dorefa-net: Training low bitwidth convolutional neural networks with low bitwidth gradients;zhou,2016

2. Introducing 2nd generation IPU systems for AI at scale;toon,2020

3. Intel deep neural network library,0

4. Pytorch;paszke,2020

5. Neural graph collaborative filtering algorithm in pytorch,0

Cited by 7 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. A Stochastic Rounding-Enabled Low-Precision Floating-Point MAC for DNN Training;2024 Design, Automation & Test in Europe Conference & Exhibition (DATE);2024-03-25

2. On the resilience of representative and novel data formats in CNNs;2023 IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFT);2023-10-03

3. Generating Posit-Based Accelerators With High-Level Synthesis;IEEE Transactions on Circuits and Systems I: Regular Papers;2023-10

5. A Scalable BFloat16 Dot-Product Architecture for Deep Learning;Proceedings of the Great Lakes Symposium on VLSI 2023;2023-06-05