notesum.ai

Published at November 15

AMXFP4: Taming Activation Outliers with Asymmetric Microscaling Floating-Point for 4-bit LLM Inference

cs.AI

Released Date: November 15, 2024

Authors: Janghwan Lee1, Jiwoong Park1, Jinseok Kim2, Yongjik Kim2, Jungju Oh2, Jinwook Oh2, Jungwook Choi1

Aff.: 1Department of Electronic Engineering, Hanyang University, Seoul, Republic of Korea; 2Rebellions Inc., Republic of Korea

Arxiv: http://arxiv.org/abs/2411.09909v1