Wide Open: NVIDIA Accelerates Inference on Meta Llama 3
原文英文，约600词，阅读约需2分钟。发表于：。NVIDIA today announced optimizations across all its platforms to accelerate Meta Llama 3, the latest generation of the large language model (LLM). The open model combined with NVIDIA accelerated computing equips developers, researchers and businesses to innovate responsibly across a wide variety of applications. Trained on NVIDIA AI Meta engineers trained Llama 3 on computer Read Article
Meta和NVIDIA合作开发了Meta Llama 3，这是一个优化在NVIDIA GPU上运行的开放式大型语言模型（LLM）。该模型在一个由24,576个NVIDIA H100 Tensor Core GPU组成的计算机集群上进行了训练，并可在云端、数据中心、边缘和个人电脑上使用。企业可以使用NVIDIA NeMo对Llama 3进行数据微调，并使用NVIDIA Triton Inference Server部署自定义模型。Llama 3还可在NVIDIA Jetson Orin上运行，用于机器人和边缘计算设备。NVIDIA致力于优化社区软件并推动AI透明度。
