在 NVIDIA DGX Spark 上提升 MoE 推理性能的 3 种方法
3 Ways to Boost MoE Inference on NVIDIA DGX Spark
NVIDIA 技术博客给出在 DGX Spark 上加速 MoE 推理的 3 种方法。对 Inclusion AI 的 Ling-3.0-flash 做 batch-size-1 评测时，性能剖析显示解码受内存带宽限制，读取模型权重主导了执行时间，MoE 计算与输出层投影两个环节尤为突出。
Generating tokens faster starts with identifying what limits inference performance. In the batch-size-1 evaluation of Ling-3.0-flash from Inclusion AI on NVIDIA...
Generating tokens faster starts with identifying what limits inference performance. In the batch-size-1 evaluation of Ling-3.0-flash from Inclusion AI on NVIDIA DGX Spark, profiling showed that decoding was memory-bound: reading model weights drove execution time. Two operations stood out—the mixture-of-experts (MoE) computation and the output-layer projection that produces scores across the model’…
来源：NVIDIA Technical Blog · developer.nvidia.com