Session

· Contributors · Organizations · Search Program · My Agenda

Research Manuscript: Fantastic AI System Improvements and Where to Find Them

Session Chairs

Lingchuan Meng

Amazon

Jinho Lee

Seoul National University

Event TypeResearch Manuscript

Topics

Keywords

TimeWednesday, July 12th10:30am - 12:00pm PDT

Location3012, 3rd Floor

DescriptionWe all agree that AI efficiency improvements are important and the keys to unlock more use cases. But the question is where to find them in the system. This session presents paper that improves inference efficiency by optimizing different parts of an AI system, ranging from non-volatile storage, memory usage, compute instruction packing, and to power management.

Presentations

10:30am - 10:40am PDT	Lightning Talk: Emerging Hardware Technologies and 3D System Integration for Ubiquitous Machine Intelligence Author Haitong Li
10:40am - 10:55am PDT	Intermittent-Aware Neural Network Pruning Authors Chih-Chia Lin Chia-Yin Liu Chih-Hsuan Yen Tei-Wei Kuo Pi-Cheng Hsiu
10:55am - 11:10am PDT	Occamy: Memory-efficient GPU Compiler for DNN Inference Authors Jaeho Lee Shinnung Jeong Seungbin Song Kunwoo Kim Heelim Choi Youngsok Kim Hanjun Kim
11:10am - 11:25am PDT	Uint-Packing: A Unsigned Full-Integer DSP Packing Method Multiples Your DNN Accelerator Performance Authors Jingwei Zhang Meng Zhang Xinye Cao Guoqing Li
11:25am - 11:40am PDT	Condense: A Framework for Device and Frequency Adaptive Neural Network Models on the Edge Authors Yifan Gong PU ZHAO zheng zhan Yushu Wu Chao Wu Zhenglun Kong Minghai Qin Caiwen Ding Yanzhi Wang