CRISP：面向边缘服务器 MoE 推理的剪枝与调度协同设计
CRISP: Co-designing Pruning and Scheduling for Efficient MoE Inference on Edge Servers
MoE 模型跑在边缘设备上会被专家权重搬运拖慢，这篇论文提出 CRISP，在 Mixtral-8x7B 上把吞吐拉到 2.1 倍，精度只掉 3.6%，做端侧部署的可以看看。
CRISP 是一个面向内存受限边缘平台的 MoE 推理加速框架，核心思路是协同设计剪枝与运行时调度。它按重要性重排神经元，使单个存储的专家支持多种执行宽度，并根据频率把专家组合分为热、温、冷三档分配差异化执行。在 Mixtral-8x7B 和 DeepSeek-MoE-16B 两个模型上测试，CRISP 在相同内存预算下吞吐量提升最高 2.1 倍，相对全宽推理的准确率损失限制在 3.6%。
CRISP: Co-designing Pruning and Scheduling for Efficient MoE Inference on Edge Servers
Mixture-of-Experts (MoE) inference on memory-constrained edge platforms is often dominated by expert-weight movement rather than computation. Although each token activates only a small subset of experts, routing is input-dependent, so all experts must remain accessible and memory movement dominates latency. Existing approaches such as uniform pruning and expert dropping overlook two properties of MoE execution: expert combinations are highly skewed in frequency, and experts contribute asymmetrically within each combination. We present CRISP, a combination-aware framework that co-designs pruning and runtime scheduling for efficient MoE serving on memory-constrained edge platforms. CRISP reorders neurons by importance so one stored expert can support multiple execution widths, assigns differentiated execution to hot, warm, and cold combinations based on frequency and expert role, and exploits heterogeneous widths to overlap expert loading with computation. Across Mixtral-8x7B and DeepSeek-MoE-16B, CRISP improves throughput by up to $2.1\times$ at the same memory budget while limiting accuracy loss to 3.6\% relative to full-width inference. Our results highlight that efficient edge MoE serving requires jointly optimizing which expert capacity is preserved and how expert execution is scheduled.
