Session

· Contributors · Organizations · Search Program · Flagged · Happening Now

Research Manuscript: The Next Step to Efficient AI: Number Formats, Quantization and Beyond

Session Chairs

DescriptionThe success of AI has been accompanied by an astronomical increase in computational requirements, while at the same time there is a drive to deploy complex AI algorithms on resource-constrained edge devices. The papers in this session explore new number formats, quantization techniques, and hardware-friendly algorithm design for deep neural networks including Large Language Models.

Event TypeResearch Manuscript

TimeTuesday, June 251:30pm - 3:00pm PDT

Location3001, 3rd Floor

Topics

Keywords

Presentations

1:30pm - 1:48pm PDT	Algorithm-Hardware Co-Design of Distribution-Aware Logarithmic-Posit Encodings for Efficient DNN Inference Authors Akshat Ramachandran Zishen Wan Geonhwa Jeong John Gustafson Tushar Krishna
1:48pm - 2:06pm PDT	Genetic Quantization-Aware Approximation for Non-Linear Operations in Transformers Authors Pingcheng Dong Yonghao Tan Dong Zhang Tianwei Ni Xuejiao Liu Yu Liu Peng Luo Luhong Liang Shi-Yang Liu Xijie Huang Huaiyu Zhu Yun Pan Fengwei An Kwang-Ting Cheng
2:06pm - 2:24pm PDT	RL-PTQ: RL-based Mixed Precision Quantization for Hybrid Vision Transformers Authors Eunji Kwon Minxuan Zhou Weihong Xu Tajana Rosing Seokhyeong Kang
2:24pm - 2:42pm PDT	APTQ: Attention-aware Post-Training Mixed-Precision Quantization for Large Language Models Authors Ziyi Guan Hantao Huang Yupeng Su Hong Huang Ngai Wong Hao Yu
2:42pm - 3:00pm PDT	EDGE-LLM: Enabling Efficient Large Language Model Adaptation on Edge Devices via Unified Compression and Adaptive Layer Voting Authors Zhongzhi Yu Zheng Wang Yuhan Li Ruijie Gao Xiaoya Zhou Sreenidhi Reddy Bommu Yang (Katie) Zhao Yingyan (Celine) Lin

DAC 2024