PulseAugur
中
实时 06:41:43
English(EN) ExFold: Unified Expert Folding for Training-Free MoE Prefill-Decode Acceleration

ExFold框架以最小的质量损失加速MoE模型推理

作者 PulseAugur 编辑部 · [1 个来源] ·

研究人员推出ExFold，一个旨在加速混合专家（MoE）模型推理速度的新型框架。这种无需训练的方法通过将排除的专家的贡献映射到保留的专家上来解决MoE预填充和解码阶段的独特瓶颈。ExFold实现了显著的加速，首个标记生成时间（time-to-first-token）最多可提高1.41倍，每个输出标记生成时间（time-per-output-token）最多可提高2.45倍，同时保持约99%的原始模型质量。该框架作为vLLM的插件实现，并包含一个专门的CUDA内核以实现高效的专家折叠。 AI

影响 加速MoE模型推理，可能实现更大语言模型的更快、更高效的部署。

排序理由 该集群描述了一篇详细介绍用于加速AI模型推理的新型框架的新研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

ExFold框架以最小的质量损失加速MoE模型推理

本文如何被排名

Signal score
28 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该集群描述了一篇详细介绍用于加速AI模型推理的新型框架的新研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.LG TIER_1 English(EN) · Juntong Wu, Yifei Liu, Junyi Chen, Siqi Fan, Chaoran Feng, Minghao Li, Liujie Zhang, Weihang Chen, Li Yuan ·

    ExFold：统一专家折叠，实现无需训练的MoE预填充-解码加速

    arXiv:2608.24938v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models scale capacity for strong quality while keeping per-token compute bounded through sparse expert activation. Yet low-latency MoE serving is increasingly challenging, because it spans two inference phas…

相关实体

相关话题