Výsledky vyhledávání

Akademický článek

Study on the influence of reed fiber length and moisture content on the mechanical properties of concrete

Autor: Wang Xianpeng, Yu Haoxuan, Yu Kequan, Kovshar Sergey, Leonovich Sergey

Publikováno v: E3S Web of Conferences, Vol 563, p 02012 (2024)

This paper adopts a contrastive experimental method to design and prepare reed fiber concrete. It discusses the influence of reed fiber content (2%), fiber length (1.5 cm for short reed fiber and 3 cm for long reed fiber), and moisture content (0% fo

Externí odkaz: https://doaj.org/article/0b6c5b34a311483cb55a69c81e398a9e

Zobrazit plný text záznamu

Report

Missile: Fine-Grained, Hardware-Level GPU Resource Isolation for Multi-Tenant DNN Inference

Autor: Zhang, Yongkang, Yu, Haoxuan, Han, Chenxia, Wang, Cheng, Lu, Baotong, Li, Yang, Chu, Xiaowen, Li, Huaicheng

Colocating high-priority, latency-sensitive (LS) and low-priority, best-effort (BE) DNN inference services reduces the total cost of ownership (TCO) of GPU clusters. Limited by bottlenecks such as VRAM channel conflicts and PCIe bus contentions, exis

Externí odkaz: http://arxiv.org/abs/2407.13996

Zobrazit plný text záznamu

Report

FaaSwap: SLO-Aware, GPU-Efficient Serverless Inference via Model Swapping

Autor: Yu, Minchen, Wang, Ao, Chen, Dong, Yu, Haoxuan, Luo, Xiaonan, Li, Zhuohao, Wang, Wei, Chen, Ruichuan, Nie, Dapeng, Yang, Haoran

Serverless computing has become increasingly popular for machine learning inference. However, current serverless platforms lack efficient support for GPUs, limiting their ability to deliver low-latency inference. In this paper, we propose FaaSwap, a

Externí odkaz: http://arxiv.org/abs/2306.03622

Zobrazit plný text záznamu