<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>NVFP4 on AI 实战派 · 从技术到赚钱</title><link>https://guijiagi.com/tags/nvfp4/</link><description>Recent content in NVFP4 on AI 实战派 · 从技术到赚钱</description><generator>Hugo</generator><language>zh-cn</language><copyright>本站内容采用 CC BY-NC-SA 4.0 国际许可协议授权</copyright><lastBuildDate>Thu, 08 Oct 2026 15:48:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/nvfp4/index.xml" rel="self" type="application/rss+xml"/><item><title>从 GPTQ/AWQ 到 NVFP4：2026 年为什么量化重新变成硬件原生</title><link>https://guijiagi.com/posts/2026-10-08-09-nvfp4-quantization-blackwell/</link><pubDate>Thu, 08 Oct 2026 15:48:00 +0800</pubDate><guid>https://guijiagi.com/posts/2026-10-08-09-nvfp4-quantization-blackwell/</guid><description>Blackwell 之后，FP4 从后训练量化技巧变成硬件原生格式。本文梳理 NVFP4、MXFP4 与 DeepGEMM 这条新量化路线。</description></item></channel></rss>