<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Triton on 黄文卓 | DevOps Engineer</title><link>https://socake.github.io/tags/triton/</link><description>Recent content in Triton on 黄文卓 | DevOps Engineer</description><generator>Hugo -- gohugo.io</generator><language>zh-CN</language><managingEditor>17691281867@163.com (Wenzhuo Huang)</managingEditor><webMaster>17691281867@163.com (Wenzhuo Huang)</webMaster><copyright>© 2026 Wenzhuo Huang</copyright><lastBuildDate>Sun, 22 Mar 2026 09:15:00 +0800</lastBuildDate><atom:link href="https://socake.github.io/tags/triton/index.xml" rel="self" type="application/rss+xml"/><item><title>Unsloth 高效微调实战：单卡 QLoRA 的极致性能与内部原理</title><link>https://socake.github.io/posts/unsloth-efficient-finetuning/</link><pubDate>Sun, 22 Mar 2026 09:15:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/unsloth-efficient-finetuning/</guid><description>Unsloth 用手写 Triton kernel 把单卡 LoRA 微调速度和显存压到极致。本文讲清 Unsloth 的原理、和 LLaMA Factory/TRL 的组合用法，以及真实使用的坑。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/unsloth-efficient-finetuning/featured.jpg"/></item><item><title>Triton Inference Server 生产部署：模型编排、动态批处理与多框架混部</title><link>https://socake.github.io/posts/triton-inference-server-production/</link><pubDate>Wed, 11 Mar 2026 10:00:00 +0800</pubDate><author>17691281867@163.com (Wenzhuo Huang)</author><guid>https://socake.github.io/posts/triton-inference-server-production/</guid><description>把 Triton 从一个陌生的 NVIDIA 推理服务器讲清楚：model repository、backend、动态批处理、ensemble、BLS、Python backend、生产监控和踩坑实录。</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://socake.github.io/posts/triton-inference-server-production/featured.jpg"/></item></channel></rss>