mirror of
https://github.com/NixOS/nixpkgs.git
synced 2026-07-06 17:13:24 -05:00
68 lines
1.4 KiB
Nix
68 lines
1.4 KiB
Nix
{
|
|
lib,
|
|
buildPythonPackage,
|
|
fetchFromGitHub,
|
|
# build-system
|
|
setuptools,
|
|
# dependencies
|
|
emoji,
|
|
networkx,
|
|
numpy,
|
|
peft,
|
|
platformdirs,
|
|
protobuf,
|
|
requests,
|
|
torch,
|
|
tqdm,
|
|
transformers,
|
|
udtools,
|
|
}:
|
|
|
|
buildPythonPackage (finalAttrs: {
|
|
pname = "stanza";
|
|
version = "1.12.1";
|
|
pyproject = true;
|
|
|
|
src = fetchFromGitHub {
|
|
owner = "stanfordnlp";
|
|
repo = "stanza";
|
|
tag = "v${finalAttrs.version}";
|
|
hash = "sha256-e4f/Th9zCVBe05v7NI+IXiSbCfcKVZvDFAKq8ZrsQjA=";
|
|
};
|
|
|
|
build-system = [ setuptools ];
|
|
|
|
dependencies = [
|
|
emoji
|
|
networkx
|
|
numpy
|
|
peft
|
|
platformdirs
|
|
protobuf
|
|
requests
|
|
torch
|
|
tqdm
|
|
transformers
|
|
udtools
|
|
];
|
|
|
|
# Most tests require resources from the network (models). Many of the ones that do run are slow
|
|
# and some of them fail.
|
|
#
|
|
# Maintaining a list of "tests we can actually run in CI" isn't feasible, there are WAY too many
|
|
# exceptions and no useful pytest marks.
|
|
doCheck = false;
|
|
|
|
pythonImportsCheck = [ "stanza" ];
|
|
|
|
meta = {
|
|
description = "Official Stanford NLP Python Library for Many Human Languages";
|
|
homepage = "https://github.com/stanfordnlp/stanza/";
|
|
changelog = "https://github.com/stanfordnlp/stanza/releases/tag/${finalAttrs.src.tag}";
|
|
license = lib.licenses.asl20;
|
|
maintainers = with lib.maintainers; [
|
|
riotbib
|
|
Stebalien
|
|
];
|
|
};
|
|
})
|