<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Transformer on jimyag's Blog</title><link>https://jimyag.com/blog/transformer/</link><description>jimyag 的个人博客，记录 Kubernetes、KubeVirt、Kube-OVN、Firecracker、Go 与云原生工程实践。</description><generator>Hugo</generator><language>zh-CN</language><copyright>Copyright © 2019, jimyag.</copyright><lastBuildDate>Sun, 01 Nov 2026 16:59:59 +0800</lastBuildDate><atom:link href="https://jimyag.com/blog/transformer/index.xml" rel="self" type="application/rss+xml"/><item><title>现代 LLM 原理：从一次回答拆开 Token、Transformer 与训练过程</title><link>https://jimyag.com/posts/how-modern-llms-work/</link><pubDate>Thu, 29 Oct 2026 23:24:12 +0800</pubDate><guid>https://jimyag.com/posts/how-modern-llms-work/</guid><description>从一次具体的下一 Token 预测出发，逐步拆开分词、向量、注意力、前馈网络、预训练、后训练、采样、上下文窗口与 KV Cache，理解现代大语言模型是怎样训练并生成回答的。</description></item></channel></rss>