Cerebras 如何用 59 块晶圆部署 3 万亿参数 Kimi K3
原标题：How we deploy Trillion-parameter frontier models on Cerebras
AI 摘要
Cerebras 在博客中说明如何将万亿参数级前沿模型部署在其晶圆级芯片 WSE-3T 上。以约 3 万亿参数的开源模型 Kimi K3 为例，团队按层类型将模型切分到多块晶圆：线性注意力层每片放两层、稠密层每片放一层并预留 KV cache，共需 59 块晶圆。晶圆间仅传递激活值，通过 RDMA over Ethernet 连接，每跳约 2 微秒，并以流水线方式同时处理多个请求，从而实现 1000-2000+ tokens/秒的推理速度。
正文节选
One of the most frequently asked questions we get is how we deploy large models on Cerebras wafer-scale chips. A single Cerebras WSE-3T wafer has 44 GB of on-chip memory, and trillion-parameter frontier models need terabytes for their weights alone. In this post, we’ll explain how we split a large frontier model across multiple WSE-3T wafers, to run multi-trillion parameter models at 1000-2000+ tokens per second. We’ll use the open source model Kimi K3, with ~3 trillion parameters, as an illustr