PulseAugur
中
实时 19:35:08
English(EN) MESH: Memory-Efficient Sinkhorn Optimization for Mixture-of-Experts Training

新的 MESH 优化器提高了 MoE 训练效率,降低了内存使用

研究人员开发了 MESH,这是一种新颖的优化技术,旨在提高混合专家(MoE)模型训练的效率。传统的内存高效优化器(如 Sinkhorn)由于 MoE 架构的梯度具有条件性和可变性,因此在 MoE 架构上存在困难。MESH 引入了一种隐藏动量 Sinkhorn 更新,可保留时间一阶矩信号,与 AdamW 相比,优化器状态内存减少约 62.5%,峰值 PyTorch CUDA 分配减少约 12.6%,同时保持具有竞争力的评估损失。 AI

影响 为大型 MoE 模型引入了一种更节省内存的训练方法,有可能降低计算成本并支持更大模型的开发。

排序理由 研究论文,详细介绍了用于 AI 模型训练的新优化方法。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

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

新的 MESH 优化器提高了 MoE 训练效率,降低了内存使用

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
研究论文,详细介绍了用于 AI 模型训练的新优化方法。[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
paper, infra
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
78 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.AI TIER_1 English(EN) · Masato Fujitake ·

    MESH:混合专家训练的内存高效Sinkhorn优化

    arXiv:2608.04407v1 Announce Type: cross Abstract: Memory-efficient matrix optimizers such as Sinkhorn gradient descent remove most AdamW optimizer state for dense Transformer matrices, but direct application to Mixture-of-Experts (MoE) training is unreliable. We study this failur…