Meta 近日披露了其最新的两个24k GPU集群的详尽信息，这些集群将专门用于训练其先进的Llama3大语言模型。这一消息在Meta的“构建Meta的GenAI基础设施”博客文章中公布，详细阐述了新集群的硬件、网络、存储、设计、性能和软件配置。新集群的构建基于2022年推出的AI Research SuperCluster (RSC)，展现了公司在人工智能研究和开发领域的持续投入。
据透露，这两个新的训练集群各自包含了24,576个Nvidia Tensor Core H100 GPU，较之早期拥有16,000个Nvidia A100 GPU的集群，规模有了显著提升。这一扩展意味着Meta现在有能力处理更大、更复杂的模型，特别是在自然语言处理、语音识别和图像生成等关键领域。Meta表示，这样的硬件升级将进一步推动生成式AI产品的创新和进步。
这些GPU集群的增强性能为Meta的研究人员提供了更强大的计算能力，有望加速AI技术的边界探索，为未来的科技发展注入新的活力。随着Llama3模型在这些新集群上的训练，Meta在人工智能领域的领先地位或将得到进一步巩固，同时也预示着更智能、更高效的AI解决方案即将面世。
英语如下：
**News Title:** “Meta Constructs 24k GPU Superclusters to Power Llama3 Language Model Training”
**Keywords:** Meta, GPU clusters, Llama3
**News Content:** Meta recently revealed detailed information about its two new 24k GPU clusters, specifically designed for training its advanced Llama3 large language model. The announcement was made in Meta’s “Building Meta’s GenAI Infrastructure” blog post, which delves into the hardware, network, storage, design, performance, and software configurations of the new clusters. These clusters are an extension of the AI Research SuperCluster (RSC) launched in 2022, demonstrating the company’s ongoing commitment to AI research and development.
It has been disclosed that each of these new training clusters comprises 24,576 Nvidia Tensor Core H100 GPUs, a significant increase from the earlier clusters with 16,000 Nvidia A100 GPUs. This expansion enables Meta to handle larger and more complex models, particularly in key areas such as natural language processing, speech recognition, and image generation. Meta asserts that such hardware upgrades will fuel innovation and advancements in generative AI products.
The enhanced performance of these GPU clusters provides Meta’s researchers with unprecedented computational power, potentially accelerating the exploration of AI’s frontiers and injecting new momentum into technological progress. As the Llama3 model undergoes training on these new clusters, Meta’s position at the forefront of AI could be further solidified, foreshadowing the emergence of smarter and more efficient AI solutions.
【来源】https://www.datacenterdynamics.com/en/news/meta-reveals-details-of-two-new-24k-gpu-ai-clusters/
Views: 2