<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>分布式推理 on PlumePHP</title><link>https://plumephp.com/tags/%E5%88%86%E5%B8%83%E5%BC%8F%E6%8E%A8%E7%90%86/</link><description>Recent content in 分布式推理 on PlumePHP</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Sun, 27 Sep 2026 09:30:00 +0800</lastBuildDate><atom:link href="https://plumephp.com/tags/%E5%88%86%E5%B8%83%E5%BC%8F%E6%8E%A8%E7%90%86/index.xml" rel="self" type="application/rss+xml"/><item><title>分布式推理与 GPU 集群调度：张量并行、流水并行与调度器实战</title><link>https://plumephp.com/ai-distributed-inference-gpu-cluster/</link><pubDate>Sun, 27 Sep 2026 09:30:00 +0800</pubDate><guid>https://plumephp.com/ai-distributed-inference-gpu-cluster/</guid><description>&lt;blockquote&gt;
&lt;p&gt;当单卡放不下一个 70B 模型，或单卡无法满足线上吞吐时，分布式推理就从&amp;quot;可选优化&amp;quot;变成&amp;quot;必选项&amp;quot;。本文系统讲解张量并行、流水线并行与专家并行三大策略的通信代价与适用场景，介绍连续批处理在集群中的落地，并深入 KServe 与 Ray Serve 两类 GPU 调度器的架构与配置。&lt;/p&gt;</description></item></channel></rss>