GGUF · 学 AI | 投机取巧

解决什么问题What problem it solves

本地跑模型的事实标准文件格式。llama.cpp 生态几乎只认它。The de facto standard file format for running models locally. The llama.cpp ecosystem accepts little else.

这是什么What it is

llama.cpp 项目定义的模型文件格式，把模型结构与量化参数打包在一起。

目前社区量化版本绝大多数产出这个格式。

The model file format defined by the llama.cpp project, packaging the model structure together with its quantisation parameters.

The vast majority of community quantisations ship in this format.

为什么需要知道Why it matters

如果你要本地跑模型，下载到的基本就是它。非 GGUF 格式往往需要额外转换步骤。If you run models locally, this is what you download. Other formats usually need an extra conversion step.

接着读什么，为什么What to read next, and why

llama.cppllama.cpp
分清文件格式与读取它的运行器。Distinguish a file format from the runtime reading it.
量化Quantization
在《一周路线》中接着读Continue in “The Full Week”

现在能做什么What you can do now

出处Sources

GGUFGGUF