Build on pythonImportsCheckPhase: ModuleNotFoundError: No module named 'diffusers'

Help
1

Hello, probably i had made a mistake somewhere in default.nix while i was packaging huggingface/diffusers.

Executing pythonImportsCheckPhase
Check whether the following modules can be imported: diffusers
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "<string>", line 1, in <lambda>
  File "/nix/store/95cxzy2hpizr23343b8bskl4yacf4b3l-python3-3.10.11/lib/python3.10/importlib/__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1004, in _find_and_load_unlocked
ModuleNotFoundError: No module named 'diffusers'

I am attaching the default.nix code.
How i run the build: nix-build -A python310Packages.diffusers inside of nixpkgs cloned repository/

2

Try disabling pythonImportsCheck and see if you can import the module after building.

3

Still can’t import diffusers

4

@collinarnett Hello, i used your repository, thank you. Maybe you know where is the problem?

5

I couldn’t make the tests work, but this worked for me

still needs some work to be submitted to nixpkgs, but hopefully this helps

{ lib
, buildPythonPackage
, fetchFromGitHub
, filelock
, huggingface-hub
, importlib-metadata
, numpy
, pillow
, regex
, requests
, accelerate
, black
  # , compel
, datasets
, flax
  # , hf-doc-builder
, isort
, jax
, jaxlib
, jinja2
  # , k-diffusion
, librosa
  # , note-seq
, parameterized
, protobuf
, pytest
, pytest-timeout
, pytest-xdist
, requests-mock
, ruff
, safetensors
, scipy
, sentencepiece
, tensorboard
, torch
, torchvision
, transformers
}:

buildPythonPackage rec {
  pname = "diffusers";
  version = "0.16.1";
  format = "setuptools";

  src = fetchFromGitHub {
    owner = "huggingface";
    repo = "diffusers";
    rev = "v${version}";
    hash = "sha256-/p+8w3YJPWj+XBAvPObIo8CcN2yjZekfV0AR3nkhUds=";
  };

  propagatedBuildInputs = [
    filelock
    huggingface-hub
    importlib-metadata
    numpy
    pillow
    regex
    requests
  ];

  passthru.optional-dependencies = {
    dev = [
      accelerate
      black
      # compel
      datasets
      flax
      # hf-doc-builder
      isort
      jax
      jaxlib
      jinja2
      # k-diffusion
      librosa
      # note-seq
      parameterized
      protobuf
      pytest
      pytest-timeout
      pytest-xdist
      requests-mock
      ruff
      safetensors
      scipy
      sentencepiece
      tensorboard
      torch
      torchvision
      transformers
    ];
    docs = [
      # hf-doc-builder
    ];
    flax = [
      flax
      jax
      jaxlib
    ];
    quality = [
      black
      # hf-doc-builder
      isort
      ruff
    ];
    test = [
      # compel
      datasets
      jinja2
      # k-diffusion
      librosa
      # note-seq
      parameterized
      pytest
      pytest-timeout
      pytest-xdist
      requests-mock
      safetensors
      scipy
      sentencepiece
      torchvision
      transformers
    ];
    torch = [
      accelerate
      torch
    ];
    training = [
      accelerate
      datasets
      jinja2
      protobuf
      tensorboard
    ];
  };

  doCheck = false;

  pythonImportsCheck = [ "diffusers" ];

  meta = with lib; {
    description = "State-of-the-art diffusion models for image and audio generation in PyTorch";
    homepage = "https://github.com/huggingface/diffusers";
    changelog = "https://github.com/huggingface/diffusers/releases/tag/${src.rev}";
    license = licenses.asl20;
    maintainers = with maintainers; [ ];
  };
}
6

Not too sure what’s wrong with your file but I just updated the hash and the version of diffusers in my repo and it I get the import to succeed in a jupyter notebook; however I’m unsure if my flake check is actually executing pythonImportsCheck

I just updated the repo if you want to check it out. Thank you for your interest as well. I initially wanted to package this for nixpkgs but honestly the process seems pretty intimidating.