vLLM 出了个新插件，把 MoE 推理的注意力层和专家层拆分部署，想分别扩缩容就方便了，支持 NVIDIA 和昇腾卡。
vLLM 项目发布实验性插件 vLLM AFD Plugin，由 Ascend、vLLM、阶跃星辰、蚂蚁集团、FastAFD 联合开发。该插件针对 MoE 模型提出 Attention-FFN Disaggregation（AFD）方案，将注意力和专家/FFN 路径作为独立服务运行，支持独立扩缩容。AFD 运行在 NVIDIA GPU 和 Ascend NPU 上，无需 fork 即可使用相同 vLLM 服务接口。
🎉 Congrats to the teams behind vLLM AFD Plugin (As…
🎉 Congrats to the teams behind vLLM AFD Plugin (Ascend & vLLM, @StepFun_ai, @AntGroup, FastAFD): a new experimental plugin under vllm-project that brings Attention-FFN Disaggregation to MoE serving.
Attention and the expert/FFN path are two very different workloads that normally share one topology. AFD runs them as separate services, so you can scale attention and experts independently. Same vLLM serving surface, no fork. @NVIDIA GPU and Ascend NPU.