FAQ#

Some frequently asked questions would be shown here. If the error you met is not contained here, please open an issue on out project abess-team/abess#issues.

Python package#

Installation failed#

1. Compliers#

First of all, please check the version of Python and GCC/MSVC. To make sure that abess package runs correctly,

  • Python 3.5 or later is required;

  • On Linux and MacOS, GCC 4.7 or later is required (support c++11);

  • On Windows, Microsoft Visual C++ (MSVC) Build Tools is required;

What's more, the newer version is recommended. So if you meet some errors, please try to update the complier first.

Besides, in Windows, you may receive an error said “error: Microsoft Visual C++ *version* is required”. To fix it, you need to check if MSVC is installed and enabled correctly. That is, to download Microsoft C++ Build Tools and install (or re-install) the "Desktop development with C++" module inside.

2. Permission#

If you receive an error said "Can't create or remove files in install directory" during the installation, this may be caused by permission denied. The step below would help with it.

  • For Linux/MacOS: run $ python setup.py install --user or $ sudo python setup.py install instead.

  • For Windows: run $ python setup.py install --user or run the command as an administrator.

3. Dependencies#

If you receive an error message including:

...
Could not find a package configuration file provided by "pybind11" with any
of the following names:

    pybind11Config.cmake
    pybind11-config.cmake*"
...

it means pybind11 has not been installed properly, and please check it again.

What's more, this situation usually happen when you install by pip install pybind11. Instead, we recommend using pip install pybind11[global] to avoid it.

Import failed?#

1. Folder name#

Make sure your working folder path is not named “abess”. If not, Python would not import the abess packages and give some errors.

Example not reproducible?#

1. Datasets#

Some examples are based on data generated by abess.datasets. Since it depends on numpy.random and it would causes differences on a different OS or CPU, you might get a different dataset and finally a different result.

R package#

  • Update Rcpp package if you encounter the following errors:

function 'Rcpp_precious_remove' not provided by package 'Rcpp'