← 研踪 PaperTrailarXiv:2611.10219 · v1
Multi-GPU MoE Inference with Adaptive Expert Offloading
首次提交 2026-11-07 05:02
版本更新 2026-11-07 05:02
雷达观测 2026-11-10 10:11
摘要
Modern Mixture-of-Experts (MoE) inference is constrained by GPU memory. This constraint is mitigated by expert offloading, which stores part of the model's weight in host memory and executes on demand. Though MoE architecture has computation sparsity, more GPUs are needed for serving increasingly larger frontier models. However, existing expert offloading systems overlook this need and fail to scale performance with additional GPUs. We present AdaOffload, a general and efficient multi-GPU hybrid expert execution framework that effectively handles heterogeneous hardware and workloads. We co-design GPU-CPU parallelism and GPU-CPU co-execution for decode and prefill with online workload observation that adapts to different hardware, workload, and batch size heterogeneity. We evaluated AdaOffload on five different platforms with varying GPU capability (L4, Pro6000, B200), number of GPUs, CPU capability, and NUMA topology on three different frontier open-weight models (284B to 753B). AdaOffload achieves $1.9\times$ to $15.5\times$ throughput speedups across different models over state-of-the-art systems.
作者备注：15 pages, including references and appendix
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-07 05:02 · 最新观测版本
Multi-GPU MoE Inference with Adaptive Expert Offloading
Modern Mixture-of-Experts (MoE) inference is constrained by GPU memory. This constraint is mitigated by expert offloading, which stores part of the model's weight in host memory and executes on demand. Though MoE architecture has computation sparsity, more GPUs are needed for serving increasingly larger frontier models. However, existing expert offloading systems overlook this need and fail to scale performance with additional GPUs. We present AdaOffload, a general and efficient multi-GPU hybrid expert execution framework that effectively handles heterogeneous hardware and workloads. We co-design GPU-CPU parallelism and GPU-CPU co-execution for decode and prefill with online workload observation that adapts to different hardware, workload, and batch size heterogeneity. We evaluated AdaOffload on five different platforms with varying GPU capability (L4, Pro6000, B200), number of GPUs, CPU capability, and NUMA topology on three different frontier open-weight models (284B to 753B). AdaOffload achieves $1.9\times$ to $15.5\times$ throughput speedups across different models over state-of-the-art systems.
查看此版本解读 →查看此版本原文 ↗