arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Wisconsin-Madison(威斯康星大学麦迪逊分校)

2026-04-27 至 2026-04-27 共收录 2
2510.25977 2026-04-27 cs.CL

NeuronMLP: Efficient LLM Inference via Singular Value Decomposition Compression and Tiling on AWS Trainium

NeuronMLP:通过AWS Trainium上的奇异值分解压缩和分块实现高效的LLM推理

Dinghong Song, Jierui Xu, Weichu Yang, Pengfei Su, Dong Li

机构 * University of California, Merced USA University of Wisconsin, Madison USA Yotta Labs \& University of California, Merced USA University of California, Merced University of Wisconsin, Madison Yotta Labs \& University of California, Merced

AI总结 本文提出NeuronMLP,基于奇异值分解压缩和分块技术优化AWS Trainium上的LLM推理,通过定制化技术减少数据移动,提升SRAM带宽,实现1.35倍的内核加速和1.21倍的端到端推理加速。

Comments 12 pages, 8 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2504.08170 2026-04-27 quant-ph cs.LG physics.comp-ph

Efficient measurement of neutral-atom qubits with matched filters

高效测量中性原子量子比特的匹配滤波器

Robert M. Kent, Linipun Phuttitarn, Chaithanya Naik Mude, Swamit Tannu, Mark Saffman, Gregory Lafyatis, Daniel J. Gauthier

机构 * The Ohio State University, Department of Physics(俄亥俄州立大学物理系) Department of Physics, University of Wisconsin-Madison(威斯康星大学麦迪逊分校物理系) Department of Computer Sciences, University of Wisconsin-Madison(威斯康星大学麦迪逊分校计算机科学系) Infleqtion, Inc.(Infleqtion公司) ResCon Technologies, LLC(ResCon Technologies LLC)

AI总结 本文提出两种简单可扩展的机器学习算法,用于解决中性原子量子处理器的读取串扰问题,通过局部模型和邻近量子比特信息减少误差,提升测量精度。

Journal ref Phys. Rev. Applied 25, 044048 (2026)

详情

展开后加载摘要…

URL PDF HTML 收藏