UniMoMo：基于专家合并的推荐系统MoE加速方法
原标题：UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models
AI 摘要
UniMoMo 是一种针对大型推荐系统中混合专家（MoE）模型的训练后压缩框架，通过功能相似性分组和层自适应保护，将训练好的 MoE 模型压缩为更小的标准 MoE 检查点，在保持精度的同时加速推理。实验表明，在 Amazon Beauty、KuaiRec 和 TenRec 数据集上，将专家数从 8 压缩到 4 时，NDCG@10 保持 99.92%–102.30%，A100 推理速度提升 1.28–1.63 倍；更激进的 8 到 2 专家设置下，速度提升可达 2.21 倍。该方法无需额外的在线模块，压缩后的模型仍为标准 top-k MoE，便于部署。
正文节选
UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models Abstract UniMoMo compresses trained recommendation mixture-of-experts models into smaller standard MoE checkpoints via functional similarity grouping and layer-adaptive protection, preserving accuracy while accelerating inference. Sparse mixture-of-experts (MoE) layers expand recommendation capacity through conditional computation, yet a trained checkpoint still stores and routes over its full expert bank. We study a