<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>AI on kj’s blog</title><link>https://kfjie.me/en/tags/ai/</link><description>Recent content from kj’s blog</description><generator>Hugo</generator><language>en</language><managingEditor>None (Kfjie)</managingEditor><webMaster>None (Kfjie)</webMaster><copyright>All articles on this blog are licensed under the BY-NC-SA license agreement unless otherwise stated. Please indicate the source when reprinting!</copyright><lastBuildDate>Thu, 29 Oct 2026 00:00:00 +0800</lastBuildDate><atom:link href="https://kfjie.me/en/tags/ai/index.xml" rel="self" type="application/rss+xml"/><item><title>How to Fine-Tune a Video Model with a 12GB Graphics Card: A Log of My ZeRO-3 Multi-GPU Experiment</title><link>https://kfjie.me/en/post/zero3-multigpu-debugging/</link><pubDate>Thu, 29 Oct 2026 00:00:00 +0800</pubDate><author>None (Kfjie)</author><guid>https://kfjie.me/en/post/zero3-multigpu-debugging/</guid><description>
<![CDATA[<h1>How to Fine-Tune a Video Model with a 12GB Graphics Card: A Log of My ZeRO-3 Multi-GPU Experiment</h1><p>Author: Kfjie(None)</p>
        
          <p>This post documents the process of fine-tuning a video model on the server that I worked on earlier.</p>
<p>We&rsquo;re using Video-ORA. The task is to have the model watch a video, output the time intervals worth keeping, and then pass that information to a subsequent program for editing. I&rsquo;m working on SFT, which involves supervised fine-tuning using the video and corresponding answers to train a small number of additional parameters using LoRA.</p>
        
        <hr><p>Published on 2026-10-29 at <a href='https://kfjie.me/'>kj’s blog</a>, last modified on 2026-10-29</p>]]></description><category>Notes</category></item><item><title>Share the Large Model with Your Family: Connect Codex to WeChat</title><link>https://kfjie.me/en/post/weixin-household-gateway/</link><pubDate>Thu, 21 May 2026 00:00:00 +0800</pubDate><author>None (Kfjie)</author><guid>https://kfjie.me/en/post/weixin-household-gateway/</guid><description>
<![CDATA[<h1>Share the Large Model with Your Family: Connect Codex to WeChat</h1><p>Author: Kfjie(None)</p>
        
          <p>This post is meant to document a project I’ve almost finished working on recently:<code>weixin-household-gateway</code>。</p>
<p>Project Address:<br>
<a href="https://github.com/thekfjie/weixin-household-gateway">github.com/thekfjie/weixin-household-gateway</a></p>
<p>Simply put, it’s a home gateway that integrates the Codex/AI backend into WeChat.<br>
It’s not a web-based agent, nor is it simply a chatbot; rather, it’s a WeChat gateway deployed on my own server: I can use it within WeChat to manage servers, code, files, and attachments, and I can also let my family members use it to have casual AI conversations in a relatively secure way.</p>
        
        <hr><p>Published on 2026-05-21 at <a href='https://kfjie.me/'>kj’s blog</a>, last modified on 2026-05-21</p>]]></description><category>Project</category></item></channel></rss>