Category I: Bridging the Gap Between AI/ML Computing Demands and Today's Capabilities

第一类：缩小 AI/ML 计算需求与当今能力之间的差距

• 批准号: 2320345
• 负责人: William Gropp
• 金额: $ 1000万
• 依托单位: University of Illinois at Urbana-Champaign
• 依托单位国家: 美国
• 项目类别: Cooperative Agreement
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-01 至 2024-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2320345&HistoricalAwards=false
• 关键词: Category; Bridging; Gap; Between; AI

Advances in computing hardware and in artificial intelligence (AI) research have led to AI-focused systems that have transformed the landscape of computational research and of society itself. Much of today's AI research relies on access to large volumes of data and advanced computational power, which are often unavailable to researchers not located at well-resourced technology companies and universities. This divide limits the ability of researchers to leverage AI to tackle the big challenges in our society. It further constrains the diversity of researchers and the breadth of ideas incorporated into AI innovations, thereby contributing to embedded biases and other systemic inequalities found in AI systems today. The National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign will deploy and operate DeltaAI, an advanced computing and data resource that will be a companion system to NCSA’s Delta system. In doing so, DeltaAI will greatly expand the AI-focused computing capacity available within the NSF-funded advanced computing ecosystem. The DeltaAI system builds upon the efforts and successes of NCSA’s Delta system in a way that is seamless and leverages the existing NSF investment in that infrastructure. This approach is achieved by deploying DeltaAI using similar system and storage hardware, and the same high-speed interconnect. A vast array of next-generation graphics processors (GPUs) delivers a technological leap in AI computing capability that will make DeltaAI the keystone in the NSF AI computing portfolio for years to come. Advanced and modern web-based interfaces will make the resource more usable by the growing community of research domains employing AI methods in their research, and significant computing capacity will make more resources available to a more diverse group of researchers.The DeltaAI system features a large and uniform pool of compute nodes that will enable advanced AI-based research, from single-node tasks to massively parallel AI codes. The system’s use of larger memory configuration GPUs will make new types of AI codes feasible that could not previously be undertaken. Taking advantage of next-generation NVIDIA graphics processors, DeltaAI builds on Delta by expanding what is already the most performant GPU computing resource in the National Science Foundation portfolio. The compute elements of DeltaAI will provide more than 300 next-generation NVIDIA graphics processors delivering over 600 petaflops of half-precision floating point computing, distributed across an advanced network interconnect for application communications and access to an innovative, flash-based storage subsystem. Expanding the capacity and performance of the Delta storage subsystem, the DeltaAI storage environment will further transform large scale data-enabled and AI-based research workloads.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.

计算硬件和人工智能(AI)研究的进步导致了专注于AI的系统，这些系统已经改变了计算研究和社会本身的格局。如今的人工智能研究在很大程度上依赖于访问大量数据和先进的计算能力，而这些往往是不在资源充足的科技公司和大学的研究人员所无法获得的。这种鸿沟限制了研究人员利用人工智能应对我们社会中的重大挑战的能力。它进一步限制了研究人员的多样性和纳入人工智能创新的想法的广度，从而助长了今天人工智能系统中发现的嵌入偏见和其他系统性不平等。位于伊利诺伊大学厄巴纳-香槟分校的国家超级计算应用中心(NCSA)将部署和运行DeltaAI，这是一种先进的计算和数据资源，将成为NCSA的Delta系统的配套系统。通过这样做，DeltaAI将极大地扩展NSF资助的高级计算生态系统中可用的以人工智能为重点的计算能力。DeltaAI系统以NCSA Delta系统的努力和成功为基础，以一种无缝的方式构建，并利用了NSF在该基础设施上的现有投资。这种方法是通过使用类似的系统和存储硬件以及相同的高速互连来部署DeltaAI来实现的。大量的下一代图形处理器(GPU)带来了AI计算能力的技术飞跃，这将使DeltaAI成为未来几年NSF AI计算产品组合中的基石。先进和现代的基于网络的界面将使资源更容易被越来越多的研究领域使用人工智能方法进行研究，而强大的计算能力将使更多的资源可用于更多样化的研究群体。DeltaAI系统具有大量统一的计算节点池，将支持从单节点任务到大规模并行AI代码的高级基于AI的研究。该系统使用更大内存配置的GPU将使新类型的AI代码成为可能，这是以前无法进行的。利用下一代NVIDIA图形处理器，DeltaAI通过扩展已经是国家科学基金会产品组合中最高性能的GPU计算资源，建立在Delta的基础上。DeltaAI的计算元素将提供300多个下一代NVIDIA图形处理器，提供超过600千万亿次半精度浮点计算，分布在先进的网络互联中，用于应用程序通信和访问创新的基于闪存的存储子系统。通过扩展Delta存储子系统的容量和性能，DeltaAI存储环境将进一步转变大规模数据启用和基于AI的研究工作负载。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。