With Shared Microexponents, A Little Shifting Goes a Long Way
Bita Darvish Rouhani
Ritchie Zhao
V. Elango
Rasoul Shafipour
Mathew Hall
Maral Mesmakhosroshahi
Ankit More
Levi Melnick
Maximilian Golub
G. Varatkar
Lai Shao
Gaurav Kolhe
Dimitry Melts
Jasmine Klar
Renee L'Heureux
Matt Perry
Doug Burger
Eric S. Chung
Zhaoxia Deng
S. Naghshineh
Jongsoo Park
Maxim Naumov

Abstract
This paper introduces Block Data Representations (BDR), a framework for exploring and evaluating a wide spectrum of narrow-precision formats for deep learning. It enables comparison of popular quantization standards, and through BDR, new formats based on shared microexponents (MX) are identified, which outperform other state-of-the-art quantization approaches, including narrow-precision floating-point and block floating-point. MX utilizes multiple levels of quantization scaling with ultra-fine scaling factors based on shared microexponents in the hardware. The effectiveness of MX is demonstrated on real-world models including large-scale generative pretraining and inferencing, and production-scale recommendation systems.
View on arXivComments on this paper