NVIDIA AI

NVIDIA AI

计算机硬件制造业

Santa Clara,CA 1,084,556 位关注者

关于我们

Preventing disease. Building smart cities. Revolutionizing analytics. These are just a few things happening today where AI initiatives are creating real results for businesses.

网站
https://nvda.ws/2nfcPK3
所属行业
计算机硬件制造业
规模
超过 10,001 人
总部
Santa Clara,CA

动态

  • 查看NVIDIA AI的公司主页,图片

    1,084,556 位关注者

    Through the NAIRR pilot, researchers at the University of Michigan, are utilizing this technology to accelerate material discoveries essential for energy storage and transportation electrification. This initiative, supported by the National Science Foundation (NSF) and other agencies, offers high-performance computing resources to enhance AI research and innovation. We are pleased to be providing DGX Cloud to researchers nationwide through this pilot.

    查看National Science Foundation (NSF)的公司主页,图片

    271,446 位关注者

    Since the NSF-led National Artificial Intelligence Research Resource (NAIRR) pilot launched in January 2024, the initiative has been laying down the foundation for an AI research ecosystem where ideas and innovation from a diverse pool of talent can thrive and benefit the nation. “The need for AI infrastructure, research and education is only going to increase. In providing access to top-tier AI resources and expertise, the government, industry and nonprofit partners of the NAIRR pilot are helping our nation’s researchers and educators develop AI for the greater good and train our rapidly expanding AI workforce. NAIRR is meeting people where they are, regardless of geography or background, channeling the United States’ unique diversity into innovations and discoveries to advance a trustworthy global AI ecosystem and drive international standards, ensuring U.S. AI leadership for decades to come,” said NSF Director Sethuraman Panchanathan. Through collaborative partnerships, the pilot has awarded more than 150 resource awards since its inception. In only a few months, the impacts from the initial round of award recipients are already emerging, demonstrating the potential groundbreaking impact that a fully implemented NAIRR could have on research, education and society. https://bit.ly/3Uf802e

    • Graphic post of NSF's National Artificial Intelligence Research Resource pilot awards map in the United States.
  • NVIDIA AI转发了

    查看Mike Sievert的档案,图片
    Mike Sievert Mike Sievert是领英影响力人物

    Chief Executive Officer at T-Mobile

    A key element of T-Mobile's strategy is extending our network leadership while transforming the company into an #AI enabled, data-informed, digital-first organization. That includes revolutionizing network design and capabilities – for our industry and beyond. ? In this special edition of Sidekicks Conversations, Jensen Huang and I share how NVIDIA and T-Mobile are partnering with Ericsson and Nokia to create an AI-RAN Innovation Center and transform the future of mobile networks. https://lnkd.in/gKqBK9AF

  • NVIDIA AI转发了

    查看Neural Magic的公司主页,图片

    16,616 位关注者

    ?? Introducing Machete: A New Mixed-Input GEMM Kernel for NVIDIA Hopper GPUs ?? We’re excited to unveil Machete, a major step forward in high-performance LLM inference. By focusing on w4a16 mixed-input quantization, Machete reduces memory usage by ~4x, making deployments significantly more efficient in memory-bound regimes. While compute-bound performance remains in line with FP16, Machete truly excels in optimizing memory bandwidth for GPTQ-style models. ?? Key highlights of Machete: - Built on CUTLASS 3.x, utilizing wgmma tensor core instructions, overcoming limitations in compute-bound scenarios. - Weight pre-shuffling for faster shared memory loads and reduced bottlenecks in large-scale LLMs. - 128-bit shared memory loads for high throughput and further reduced latency. Optimized upconversion routines to maximize tensor core utilization by converting 4-bit elements to 16-bit efficiently. With Machete, we’ve achieved a 29% faster input and 32% faster output token throughput on Llama 3.1 70B, with a TTFT of <250ms on a single H100 GPU. And that’s not all... On a 4xH100 setup, Machete delivers a 42% throughput speedup on Llama 3.1 405B—with more optimizations on the way, including support for w4a8 FP8, AWQ, QQQ, and low-batch-size performance. ?? A huge shoutout to Lucas Wilkinson for leading the development of Machete and the team at NVIDIA AI for continual support! Special thanks to 3Blue1Brown and the Manim community for the amazing animations that helped visualize these optimizations. Read the full blog here: https://lnkd.in/ggKYbmKR #AI #LLMs #GPUs #NVIDIA #vLLM #DeepLearning #MachineLearning

    • 4bit Llama 3.1 70b on a single H100
(neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w4a16)

关联主页

相似主页