<?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>ITL on AI 实战派 · 从技术到赚钱</title><link>https://guijiagi.com/tags/itl/</link><description>Recent content in ITL on AI 实战派 · 从技术到赚钱</description><generator>Hugo</generator><language>zh-cn</language><copyright>本站内容采用 CC BY-NC-SA 4.0 国际许可协议授权</copyright><lastBuildDate>Fri, 09 Oct 2026 15:42:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/itl/index.xml" rel="self" type="application/rss+xml"/><item><title>vLLM 的 Prefill 与 Decode 分离：把首 token 延迟与 token 间延迟分治</title><link>https://guijiagi.com/posts/2026-10-09-08-vllm-pd-disaggregation/</link><pubDate>Fri, 09 Oct 2026 15:42:00 +0800</pubDate><guid>https://guijiagi.com/posts/2026-10-09-08-vllm-pd-disaggregation/</guid><description>vLLM 把计算密集的 prefill 与带宽密集的 decode 拆到不同实例，分别调优 TTFT 与 ITL。本文讲清什么时候该拆、拆了怎么调度。</description></item></channel></rss>