<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>CUDA on 筆墨碎語｜RxChi1d</title><link>https://inktrace.rxchi1d.me/tags/cuda/</link><description>Recent content in CUDA on 筆墨碎語｜RxChi1d</description><generator>Hugo -- gohugo.io</generator><language>zh-tw</language><managingEditor>rchildlin@gmail.com (RxChi1d)</managingEditor><webMaster>rchildlin@gmail.com (RxChi1d)</webMaster><copyright>© 2026 RxChi1d</copyright><lastBuildDate>Sat, 11 Jul 2026 02:06:20 +0800</lastBuildDate><atom:link href="https://inktrace.rxchi1d.me/tags/cuda/index.xml" rel="self" type="application/rss+xml"/><item><title>幫 Ornith-1.0-35B 移植 MTP：從 GGUF 手術到 Hermes Agent 接入</title><link>https://inktrace.rxchi1d.me/posts/ai-technical/ornith-1-0-35b-mtp-deployment-notes/</link><pubDate>Mon, 06 Jul 2026 21:15:09 +0800</pubDate><author>rchildlin@gmail.com (RxChi1d)</author><guid>https://inktrace.rxchi1d.me/posts/ai-technical/ornith-1-0-35b-mtp-deployment-notes/</guid><description>&lt;p&gt;本文帶你用 llama.cpp 在本地端部署 Ornith-1.0-35B、接入 Hermes Agent，最後再透過移植 MTP head 進一步提升推理速度。&lt;/p&gt;</description></item></channel></rss>