<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[Ob1_M的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/Ob1_M</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; Ob1_M]]></copyright><item><title><![CDATA[【解决llama-factory导出并量化模型时爆显存】]]></title><link>https://blog.youkuaiyun.com/Ob1_M/article/details/147456193</link><guid>https://blog.youkuaiyun.com/Ob1_M/article/details/147456193</guid><author>Ob1_M</author><pubDate>Wed, 23 Apr 2025 16:02:06 +0800</pubDate><description><![CDATA[即可，本质上这个问题是通过将模型放在cpu上加载，gpu上量化解决的。导出模型后记得改回来，万一以后能用上更大显存的服务器呢？这时候在cpu上加载模型就是浪费时间了。服务器GPU是3090Ti-24G，用llama-factory来QLoRA微调Qwen2.5-14B-Instruct之后就该导出模型了。主要原因是直接在gpu上加载了高精度模型。]]></description><category></category></item></channel></rss>