mitchmindtree
a92186244b
Adds gguf, a python module for writing the gguf format, a common format for storing machine learning model weights. I ran into a need for this when using `ollama`, and trying to use the `llama.cpp/convert.py` tool to convert a model from one format to another.
34 lines
610 B
Nix
34 lines
610 B
Nix
{ lib
|
|
, buildPythonPackage
|
|
, fetchPypi
|
|
, numpy
|
|
, poetry-core
|
|
, pythonOlder
|
|
}:
|
|
buildPythonPackage rec {
|
|
pname = "gguf";
|
|
version = "0.6.0";
|
|
format = "pyproject";
|
|
|
|
disabled = pythonOlder "3.7";
|
|
|
|
src = fetchPypi {
|
|
inherit pname version;
|
|
hash = "sha256-suIuq6KhBsGtFIGGoUrZ8pxCk1Fob+nXzhbfOaBgfmU=";
|
|
};
|
|
|
|
dependencies = [
|
|
numpy
|
|
poetry-core
|
|
];
|
|
|
|
doCheck = false;
|
|
|
|
meta = with lib; {
|
|
description = "A module for writing binary files in the GGUF format";
|
|
homepage = "https://ggml.ai/";
|
|
license = licenses.mit;
|
|
maintainers = with maintainers; [ mitchmindtree ];
|
|
};
|
|
}
|