新闻
Setting a World Record for MoE Pre-Training on NVIDIA GB300 NVL72
NVIDIA Developer · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- NVIDIA GB300 NVL72 achieved 1,648 TFLOPs per GPU pre-training DeepSeek-V3 671B, a world record for mixture-of-experts model training.
- 为何重要
- Matters for engineers scaling large language models who need to understand communication bottlenecks and co-designed hardware-software performance gains.
- 注意
- Results use optimized software stacks; real-world performance depends on framework choice, model architecture, and whether your workload matches DeepSeek-V3 characteristics.
收听本摘要
- token
- mixture of experts
- deepseek
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。