<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Open-Source on Wissen Xue Blog</title>
    <link>https://blog.cdnsre.top/tags/open-source/</link>
    <description>Recent content in Open-Source on Wissen Xue Blog</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Tue, 07 Jul 2026 14:25:04 +0800</lastBuildDate>
    <atom:link href="https://blog.cdnsre.top/tags/open-source/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Understanding Open-Source Large Language Models: From Hugging Face to Ollama</title>
      <link>https://blog.cdnsre.top/posts/open-source-llm/</link>
      <pubDate>Tue, 07 Jul 2026 14:25:04 +0800</pubDate>
      <guid>https://blog.cdnsre.top/posts/open-source-llm/</guid>
      <description>&lt;p&gt;&lt;em&gt;A detailed summary for learning and teaching LLM&lt;/em&gt;&lt;/p&gt;&#xA;&lt;hr&gt;&#xA;&lt;h1 id=&#34;1-what-do-we-actually-download-from-hugging-face&#34;&gt;1. What do we actually download from Hugging Face?&lt;/h1&gt;&#xA;&lt;p&gt;One of the biggest misconceptions is that downloading a Large Language Model (LLM) means downloading its training data.&lt;/p&gt;&#xA;&lt;p&gt;&lt;strong&gt;This is NOT true.&lt;/strong&gt;&lt;/p&gt;&#xA;&lt;p&gt;When you download an open-source model from Hugging Face, you are primarily downloading &lt;strong&gt;a trained neural network&lt;/strong&gt;, not the data that trained it.&lt;/p&gt;&#xA;&lt;p&gt;Think of it like this:&lt;/p&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;Training data = millions or billions of books the student has read&lt;/li&gt;&#xA;&lt;li&gt;Model weights = everything the student has learned&lt;/li&gt;&#xA;&lt;li&gt;Source code = the student&amp;rsquo;s brain architecture&lt;/li&gt;&#xA;&lt;li&gt;Tokenizer = the student&amp;rsquo;s dictionary&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;p&gt;The books are &lt;strong&gt;not&lt;/strong&gt; included.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
