<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>AllToAll on Hellokitty&#39;s Blog</title>
    <link>https://yangyang233333.github.io/tags/alltoall/</link>
    <description>Recent content in AllToAll on Hellokitty&#39;s Blog</description>
    <generator>Hugo</generator>
    <language>zh-CN</language>
    <lastBuildDate>Wed, 02 Sep 2026 11:29:42 +0800</lastBuildDate>
    <atom:link href="https://yangyang233333.github.io/tags/alltoall/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>分布式 AI 训练通信原语：AllReduce、AllGather、ReduceScatter 与 AllToAll</title>
      <link>https://yangyang233333.github.io/posts/distributed-training-collective-communication/</link>
      <pubDate>Wed, 02 Sep 2026 11:29:42 +0800</pubDate>
      <guid>https://yangyang233333.github.io/posts/distributed-training-collective-communication/</guid>
      <description>从数据流、张量形状、通信量和训练场景出发，系统解释 Broadcast、Reduce、Gather、Scatter、AllReduce、AllGather、ReduceScatter 与 AllToAll，并给出可运行的 torch.distributed 示例。</description>
    </item>
  </channel>
</rss>
