arXiv · cs.AI · Yuseon Choi, Jingu Lee, Jungjun Oh, Sunjoo Whang, Byeongcheol Kim, Minsung Kim, Junhyuk Lee, Hoi-Jun Yoo, Sangjin Kim · 5 小时前模型评分31
ELMoE-3D：利用 MoE 内在弹性实现混合键合自推测解码的本地部署服务
ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving
导读
ELMoE-3D 提出一种基于混合键合（HB，Hybrid Bonding）的软硬件协同设计框架，把缓存加速与推测解码（SD，Speculative Decoding）统一起来，在 1–16 的批大小范围内平均提速 6.6 倍、能效提升 4.4 倍。
分类置信度
中 · 待人工复核
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
事件类型：研究发布
证据摘录
- “We propose ELMoE-3D, a hybrid-bonding (HB)-based HW-SW co-designed framework that unifies cache-based acceleration and speculative decoding to offer overall speedup across batch sizes.”
- “On our 3D-stacked hardware, ELMoE-3D achieves an average $6.6\times$ speedup and $4.4\times$ energy efficiency gain over naive MoE serving on xPU across batch sizes 1--16, and delivers $2.2\times$ speedup and $1.4\times$ energy efficiency gain over prior accelerator baselines.”
来源：arXiv · cs.AI · arxiv.org