pkgs/development/python-modules/openai-whisper/default.nix


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64

{ lib
, fetchFromGitHub
, buildPythonPackage

# propagates
, numpy
, torch
, tqdm
, more-itertools
, transformers
, ffmpeg-python

# tests
, pytestCheckHook
}:

buildPythonPackage rec {
  pname = "whisper";
  version = "unstable-2022-09-23";
  format = "setuptools";

  src = fetchFromGitHub {
    owner = "openai";
    repo = pname;
    rev = "8cf36f3508c9acd341a45eb2364239a3d81458b9";
    hash = "sha256-2RH8eM/SezqFJltelv5AjQEGpqXm980u57vrlkTEUvQ=";
  };

  postPatch = ''
    # Rely on the ffmpeg path already patched into the ffmpeg-python library
    substituteInPlace whisper/audio.py \
      --replace 'run(cmd="ffmpeg",' 'run('
  '';

  propagatedBuildInputs = [
    numpy
    torch
    tqdm
    more-itertools
    transformers
    ffmpeg-python
  ];

  preCheck = ''
    export HOME=$TMPDIR
  '';

  checkInputs = [
    pytestCheckHook
  ];

  disabledTests = [
    # requires network access to download models
    "test_transcribe"
  ];

  meta = with lib; {
    description = "General-purpose speech recognition model";
    homepage = "https://github.com/openai/whisper";
    license = licenses.mit;
    maintainers = with maintainers; [ hexa ];
  };
}