近日，Meta在官方博客中详细介绍了其两个全新的24k GPU集群，这些集群将用于训练旗下先进的大语言模型Llama 3。博文《构建Meta的GenAI基础设施》深入阐述了这两个集群的硬件配置、网络架构、存储方案、设计理念、性能指标及其软件支持。据悉，该集群的建立基于Meta于2022年推出的AI Research SuperCluster (RSC)。
新发布的集群旨在推动自然语言处理、语音识别和图像生成等领域的人工智能研究与开发。每个集群都配备了24,576个NVIDIA Tensor Core H100 GPU，相比之前含有16,000个NVIDIA A100 GPU的集群，这一数量有了显著提升。Meta表示，这一增长使得集群能够支撑起更大、更复杂的AI模型训练，从而为生成式AI产品的研发进步提供了强有力的计算支持。
此次Meta推出的24k GPU集群，不仅体现了公司在AI领域的持续投入和创新，也显示出Meta对生成式AI技术发展的重视。随着计算能力的增强，Llama 3模型有望在语言理解、生成能力等方面取得更加显著的进步，为用户提供更加丰富和高效的服务体验。
在人工智能技术迅速发展的当下，Meta的这一举措进一步推动了AI技术的实用化和商业化进程，也为学术界和工业界提供了更加先进的 research 和 development 平台。未来，随着这些集群的投入运作，预计将产出更多具有突破性的AI研究成果，助力全人类进入一个更加智能化的时代。
英语如下：
**Title:** Meta Unveils Details of New 24k GPU Clusters to Aid Llama 3 Model Training
**Keywords:** Meta’s New Clusters, Llama 3 Training, AI Research and Development
**News Content:**
### Meta Unveils Details of New 24k GPU Clusters to Accelerate Training of Llama 3 Large Language Model
Recently, Meta detailed its two new 24k GPU clusters in an official blog post, which are intended for training the company’s advanced large language model, Llama 3. The post “Building Meta’s GenAI Infrastructure” delves into the hardware configuration, network architecture, storage solutions, design philosophy, performance metrics, and software support of these two clusters. It is reported that the establishment of these clusters is based on Meta’s AI Research SuperCluster (RSC) introduced in 2022.
The newly released clusters aim to drive research and development in artificial intelligence across fields such as natural language processing, speech recognition, and image generation. Each cluster is equipped with 24,576 NVIDIA Tensor Core H100 GPUs, a significant increase from the previous cluster that contained 16,000 NVIDIA A100 GPUs. Meta stated that this growth enables the clusters to support the training of larger and more complex AI models, providing robust computational support for the research and development of generative AI products.
The introduction of Meta’s 24k GPU clusters not only reflects the company’s ongoing investment and innovation in AI but also demonstrates Meta’s emphasis on the development of generative AI technology. With enhanced computing power, the Llama 3 model is expected to achieve more significant progress in language understanding and generation capabilities, offering users richer and more efficient service experiences.
In the context of rapid development of artificial intelligence technology, Meta’s move further advances the practical and commercialization of AI technology and provides an even more advanced research and development platform for academia and industry. In the future, as these clusters come online, it is anticipated that they will yield more groundbreaking AI research results, helping to usher in an era of greater intelligence for humanity.
【来源】https://www.datacenterdynamics.com/en/news/meta-reveals-details-of-two-new-24k-gpu-ai-clusters/
Views: 2