Problem installing scArches

[roseboom2191]

I am attempting to install scArches and the install fails with the following error message

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for pandas
Failed to build pandas
ERROR: Failed to build installable wheels for some pyproject.toml based projects (pandas)
Note: you may need to restart the kernel to use updated packages.

The installation is being attempted inside an ipython kernel (af) within a conda environment (af).
I am running Python version 3.13.2 and pandas version 2.2.3.

I have already upgraded setuptools, pip and wheel.

[Koncopd]

Hi, looks like the same thing as here #265

[james-cranley]

Hi, you've linked to an article which basically suggests we pin numpy to a version < 2.

However

1. if this is your suggestion, you haven't updated the yaml accordingly

2. I have edited the yaml and I hit an Unsatisfiable error when conda fails to solve this

Installing via pip also gives an error (which others have raised here) regarding incompatibilities.

So right now I can't install your package on Linux.

It would be really helpful if you could edit the yaml and provide the community with a working version.

Also, minor points

• the yaml uses the defaults channel which attracts a licence fee for non commercial use. Would be nice to use conda-forge

• the yaml uses a Variable heading which is not supported by conda (I think it's a mamba thing?)

Anyway, thank you in advance, looking after a package is a lot of work.

[Koncopd]

Could you provide the full output of conda and pip errors? We don’t really have any strict upper bounds and almost all dependencies are pretty standard, so I wonder what is the source of the problems. So in the original post why does it even try to build pandas if the pre-built binary wheels are on pip.

[cjuigne]

Hi, I got the same problem.
I first tried using python3.11 and now with python3.10 as you suggested in others GitHub issues.
However, install works in both cases but when I try to import the package I got the same errors.
I also tested several combinations of numpy and pandas versions, none of them worked.

Using python3.10:

ValueError                                Traceback (most recent call last)
Cell In[7], line 7
      4 # Reinstall scArches to resolve binary incompatibility
      5 get_ipython().run_line_magic('pip', 'install --force-reinstall scArches')
----> 7 import scarches as sca

File ~/.venv310/lib/python3.10/site-packages/scarches/__init__.py:1
----> 1 from . import dataset, metrics, trainers, models, zenodo, plotting, utils, classifiers
      3 __author__ = ', '.join([
      4     'Mohammad Lotfollahi',
      5     'Sergei Rybakov',
      6     'Marco Wagenstetter',
      7     'Mohsen Naghipourfar',
      8 ])
     10 __email__ = ', '.join([
     11     'mohammad.lotfollahi@helmholtz-muenchen.de',
     12     'sergei.rybakov@helmholtz-muenchen.de'
     13 ])

File ~/.venv310/lib/python3.10/site-packages/scarches/dataset/__init__.py:2
      1 from .trvae._utils import label_encoder
----> 2 from .trvae.data_handling import remove_sparsity
      3 from .trvae.anndata import AnnotatedDataset as trVAEDataset
      4 from .scpoli.anndata import MultiConditionAnnotatedDataset

File ~/.venv310/lib/python3.10/site-packages/scarches/dataset/trvae/data_handling.py:1
----> 1 import scanpy as sc
      2 from scipy import sparse
      5 def remove_sparsity(adata):

File ~/.venv310/lib/python3.10/site-packages/scanpy/__init__.py:9
      5 import sys
      7 from packaging.version import Version
----> 9 from ._utils import check_versions
     10 from ._version import __version__
     12 check_versions()

File ~/.venv310/lib/python3.10/site-packages/scanpy/_utils/__init__.py:35
     33 import h5py
     34 import numpy as np
---> 35 from anndata import __version__ as anndata_version
     36 from packaging.version import Version
     37 from scipy import sparse

File ~/.venv310/lib/python3.10/site-packages/anndata/__init__.py:19
     15 if sys.version_info < (3, 11):
     16     # Backport package for exception groups
     17     import exceptiongroup  # noqa: F401
---> 19 from ._core.anndata import AnnData
     20 from ._core.merge import concat
     21 from ._core.raw import Raw

File ~/.venv310/lib/python3.10/site-packages/anndata/_core/anndata.py:18
     16 import h5py
     17 import numpy as np
---> 18 import pandas as pd
     19 from natsort import natsorted
     20 from numpy import ma

File ~/.venv310/lib/python3.10/site-packages/pandas/__init__.py:25
     22 from pandas.compat import is_numpy_dev as _is_numpy_dev  # pyright: ignore # noqa:F401
     24 try:
---> 25     from pandas._libs import hashtable as _hashtable, lib as _lib, tslib as _tslib
     26 except ImportError as _err:  # pragma: no cover
     27     _module = _err.name

File ~/.venv310/lib/python3.10/site-packages/pandas/_libs/__init__.py:13
      1 __all__ = [
      2     "NaT",
      3     "NaTType",
   (...)
      9     "Interval",
     10 ]
---> 13 from pandas._libs.interval import Interval
     14 from pandas._libs.tslibs import (
     15     NaT,
     16     NaTType,
   (...)
     21     iNaT,
     22 )

File ~/.venv310/lib/python3.10/site-packages/pandas/_libs/interval.pyx:1, in init pandas._libs.interval()

ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject```



Thank you in advance for your help.

[Koncopd]

@cjuigne what are the versions of pandas and numpy?

[james-cranley]

I installed scArches

1. Using pip (in a clean Python 3.10 conda environment)
2. Using conda and your official yaml

In both cases I get the same error as @cjuigne

>>> import session_info, numpy, pandas
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/rfs/project/rfs-iCNyzSAaucw/jc2226/.conda/envs/scArches_pip/lib/python3.10/site-packages/pandas/__init__.py", line 22, in <module>
    from pandas.compat import is_numpy_dev as _is_numpy_dev  # pyright: ignore # noqa:F401
  File "/rfs/project/rfs-iCNyzSAaucw/jc2226/.conda/envs/scArches_pip/lib/python3.10/site-packages/pandas/compat/__init__.py", line 18, in <module>
    from pandas.compat.numpy import (
  File "/rfs/project/rfs-iCNyzSAaucw/jc2226/.conda/envs/scArches_pip/lib/python3.10/site-packages/pandas/compat/numpy/__init__.py", line 4, in <module>
    from pandas.util.version import Version
  File "/rfs/project/rfs-iCNyzSAaucw/jc2226/.conda/envs/scArches_pip/lib/python3.10/site-packages/pandas/util/__init__.py", line 2, in <module>
    from pandas.util._decorators import (  # noqa:F401
  File "/rfs/project/rfs-iCNyzSAaucw/jc2226/.conda/envs/scArches_pip/lib/python3.10/site-packages/pandas/util/_decorators.py", line 14, in <module>
    from pandas._libs.properties import cache_readonly
  File "/rfs/project/rfs-iCNyzSAaucw/jc2226/.conda/envs/scArches_pip/lib/python3.10/site-packages/pandas/_libs/__init__.py", line 13, in <module>
    from pandas._libs.interval import Interval
  File "pandas/_libs/interval.pyx", line 1, in init pandas._libs.interval
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

Please @Koncopd can you look into this?

pip.log
yaml.log

[cjuigne]

I don't really know why but (after several days of difficulties) : using python 3.11.9

I did
pip install --upgrade pandas
pip install --upgrade numpy
and then
pip install --force-reinstall scanpy seaborn scipy

not sure that is the magic recipe, but it seems to work (import scarches works, I have not try to use it yet)

I have:

• pandas in 2.2.3
• numpy in 2.2.6
• cHPL==1.0.0

[dyinboisry4u]

I had the same problem, I must downgrade numpy to 1.26

[longmanz]

I had the same issue with python 3.10 or 3.11, and switching back to python 3.9 seems to avoid the dependency errors:

conda create --name scPoli python=3.9
conda activate scPoli
pip install git+https://github.com/theislab/scarches.git

[aichander]

requirements.uv.txt

Heres a uv environment I built for this package today to overcome this issue.
To run,

# python=3.8
uv pip install -r requirements.uv.txt