CC* Compute: Nevada Bridge to AI-enabled Scientific & Engineering Computing (NvBAISEC)

CC* 计算：内华达通往人工智能科学的桥梁

• 批准号: 2201599
• 负责人: Alireza Tavakkoli
• 金额: $ 39.62万
• 依托单位: Board of Regents, NSHE, obo University of Nevada, Reno
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-07-01 至 2025-06-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2201599&HistoricalAwards=false
• 关键词: CC; Compute; Nevada; Bridge; AI

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).The University of Nevada-Reno aims to add a 10 node A100 Graphics Processing Unit (GPU)-based cluster to their pre-existing Central Processing Unit (CPU) cluster with the expressed aim of creating a more central facility to expand access across campus. The projects aims to meet present and anticipated Artificial Intelligence (AI)/Machine Learning (ML) workloads in various research group on campus. Twenty-first century science and engineering is being transformed by the increasing scales of research computing and data. Despite the increase in the demand for large-scale and data-centric computational resources, it is still a struggle to provide domain scientists with the necessary tools and support at campus-levels. Specifically, decentralized computing practices not only bottleneck research because of lack of scale and support, but also decouple computing from higher performance and deeper storage and networks. Accordingly, shifts in institutional cyberinfrastructure strategies are required, with the following guiding priorities: (1) improving user friendly access; (2) removing perceived barriers in the use of scalable infrastructure; and (3) building multidisciplinary communities for next-generation workflows.The University of Nevada, Reno (UNR) will add a new A100 GPU-based cluster to its pre-existing CPU cluster to introduce a new set of paradigms for interdisciplinary computing infrastructure and expand access across campus. Each A100 OnDemand node is equipped with 24-core CPUs and A100 GPU accelerators, interconnected with Infiniband switches to provide effective access to science drivers at UNR and externally through Open Science Grid. This cluster has potential to meet UNR’s present and anticipated workloads of various research groups on campus by addressing three main requirements: (1) capability to support dozens to hundreds of concurrent interactive session users through the Multi Instance GPU (MIG) capabilities of the A100 GPUs; (2) support for modern tensor core architectures to facilitate machine learning workflows; and (3) colocation on the UNR research perimeter network and adjacency to high performance storage. This project is a significant step for the under-resourced institution in computing resources and capability. The planned integration with Open Science Grid (OSG) will open a door to on-demand access to millions of hours of heterogeneous compute cycles for researchers on campus.This project is funded through the collaborative efforts of the Office of Advanced Cyberinfrastructure (OAC) and the Established Program to Stimulate Competitive Research (EPSCoR).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。内华达大学里诺分校的目标是在他们现有的中央处理单元（CPU）集群中添加一个基于10节点A100图形处理单元（GPU）的集群，目的是创建一个更集中的设施，以扩大整个校园的访问范围。这些项目旨在满足校园各个研究小组当前和预期的人工智能(AI)/机器学习（ML）工作负载。21世纪的科学和工程正在被不断增长的研究、计算和数据规模所改变。尽管对大规模和以数据为中心的计算资源的需求有所增加，但为领域科学家提供校园层面的必要工具和支持仍然是一项艰巨的任务。具体来说，去中心化计算实践不仅因为缺乏规模和支持而成为研究的瓶颈，而且还将计算与更高的性能和更深的存储和网络分离开来。因此，需要转变机构网络基础设施战略，其指导重点如下：(1)改善用户友好访问；(2)消除使用可扩展基础设施的感知障碍；(3)为下一代工作流程建立多学科社区。内华达大学里诺分校（UNR）将在其现有的CPU集群基础上增加一个新的基于A100 gpu的集群，为跨学科计算基础设施引入一套新的范例，并扩大整个校园的访问。每个A100 OnDemand节点配备24核cpu和A100 GPU加速器，通过Infiniband交换机互联，通过开放科学网格提供对UNR和外部科学驱动程序的有效访问。通过解决三个主要需求，该集群有潜力满足UNR目前和预期的校园各种研究小组的工作负载：(1)通过A100 GPU的多实例GPU （MIG）功能支持数十到数百个并发交互式会话用户的能力；(2)支持现代张量核心架构，方便机器学习工作流程；(3)在UNR研究周界网络上的托管和与高性能存储的邻接。这个项目对于计算资源和能力不足的机构来说是一个重要的步骤。与开放科学网格（OSG）的计划集成将为校园研究人员按需访问数百万小时的异构计算周期打开一扇门。该项目由高级网络基础设施办公室（OAC）和促进竞争研究既定计划（EPSCoR）共同资助。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Swin-SFTNet : Spatial Feature Expansion and Aggregation Using Swin Transformer for Whole Breast Micro-Mass Segmentation

• DOI: 10.1109/isbi53787.2023.10230342
• 发表时间: 2022-11
• 期刊: 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI)
• 作者: Sharif Amit Kamran;Khondker Fariha Hossain;A. Tavakkoli;G. Bebis;Salah A. Baker