<?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 敲代码的朱雀</title>
    <link>https://www.cosefinch.com/tags/transformer/</link>
    <description>Recent content in Transformer on 敲代码的朱雀</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>zh-cn</language>
    <lastBuildDate>Wed, 27 May 2026 17:33:26 +0800</lastBuildDate><atom:link href="https://www.cosefinch.com/tags/transformer/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>LLM Transformer的解码器内幕</title>
      <link>https://www.cosefinch.com/posts/llm-transformer-decoder-inside/</link>
      <pubDate>Wed, 27 May 2026 17:33:26 +0800</pubDate>
      
      <guid>https://www.cosefinch.com/posts/llm-transformer-decoder-inside/</guid>
      <description>
        
          
            LLM Transformer Decoder-Only架构，使用多头自注意力机制和前馈神经网络，经过多重迭代实现解码器
          
          
        
      </description>
    </item>
    
    <item>
      <title>LLM的Transformer架构</title>
      <link>https://www.cosefinch.com/posts/llm-transformer-architecture/</link>
      <pubDate>Wed, 27 May 2026 15:50:50 +0800</pubDate>
      
      <guid>https://www.cosefinch.com/posts/llm-transformer-architecture/</guid>
      <description>
        
          
            大语言模型使用了Transformer机制中的Decoder-Only架构，经过多轮线性/非线性变换，取最后一个向量，预测下一个Token输出
          
          
        
      </description>
    </item>
    
  </channel>
</rss>