Installing numpy for Windows 10: Importing the multiarray numpy extension module failed

Question

I am new to Python and am attempting unsuccessfully to install numpy. The current problem is an error message reading, in part, Importing the multiarray numpy extension module failed. I cannot locate this error message on the internet and am asking for help. The complete error message is at the bottom of this post.

First I thought I should provide system and version details and describe what I have done. I am running Windows 10 on a 64-bit machine and have installed Python 2.7.13 from https://www.python.org/downloads/windows/

I have also installed Anaconda, also for Python 2.7 and a 64-bit machine, downloaded from here https://www.continuum.io/downloads. Anaconda runs Python 2.7.12, rather than Python 2.7.13. When I type import numpy inside the Anaconda GUI IPython no errors or warnings appear. Nevertheless, I have the impression running numpy from inside Python 2.7.13 using IDLE (Python GUI) might be computationally faster. Python 2.7.13 appears first in my Path statement.

I downloaded numpy-1.12.0rc2+mkl-cp27-cp27m-win_amd64.whl from http://www.lfd.uci.edu/~gohlke/pythonlibs/#jpype Following suggestions from this post installing numpy from wheel format on windows ( not a supported wheel on this platform.) I renamed:

numpy-1.12.0rc2+mkl-cp27-cp27m-win_amd64.whl to: numpy-1.12.0rc2+mkl-cp27-none-win32.whl

manually by clicking on the name and editing it. The renamed file is in the folder C:\Python27

Then I used the Windows command window (cmd) to update pip typing:

python -m pip install --upgrade pip

A message said pip was already up-to-date.

Then I typed the following in the cmd window:

pip install numpy-1.12.0rc2+mkl-cp27-none-win32.whl

A message said: Successfully installed numpy-1.12.0rc2+mkl. (Prior to doing this I attempted to install the .whl file using its original name and received an error that the file was not for my platform.)

After installing numpy in Python 2.7.13 I open Python and typed:

import numpy

Here is the complete error message that was returned:

Traceback (most recent call last):
  File "<pyshell#0>", line 1, in <module>
    import numpy
  File "C:\Python27\lib\site-packages\numpy\__init__.py", line 142, in <module>
    from . import add_newdocs
  File "C:\Python27\lib\site-packages\numpy\add_newdocs.py", line 13, in <module>
    from numpy.lib import add_newdoc
  File "C:\Python27\lib\site-packages\numpy\lib\__init__.py", line 8, in <module>
    from .type_check import *
  File "C:\Python27\lib\site-packages\numpy\lib\type_check.py", line 11, in <module>
    import numpy.core.numeric as _nx
  File "C:\Python27\lib\site-packages\numpy\core\__init__.py", line 24, in <module>
    raise ImportError(msg)
ImportError: 
Importing the multiarray numpy extension module failed.  Most
likely you are trying to import a failed build of numpy.
If you're working with a numpy git repo, try `git clean -xdf` (removes all
files not under version control).  Otherwise reinstall numpy.

Numerous numpy folders and files now appear inside C:\Python27

I also tried the following two commands and received the messages shown:

>>> pip show numpy
SyntaxError: invalid syntax
>>> git clean -xdf
SyntaxError: invalid syntax
>>> 

Sorry that this post is so long, but I do not know which of the above details might be most helpful in diagnosing the problem. I have only been using Python for about 24 hours and most of that time has been spent attempting to install numpy into Python 2.7.13. Mostly I use R and know some C. I have not attempted to install any other Python libraries or packages. Perhaps I am missing dependencies required by numpy. Thank you for any advice or assistance with this problem. If I can provide additional information please let me know. Sorry if this is a duplicate.

Answer 1

This issue was discussed extensively over on https://github.com/numpy/numpy/issues/9272 with a workaround solution which worked for me. It's pertinent to the Anaconda environment I believe you have.

Quoting: tobigithub commented on Jul 23, 2017. @njsmith thank you, I just confirmed using a VM on a clean Windows 10 system that has never seen Intel MKL DLLs, numpy or conda, all things are fine and numpy works. I am just puzzled how numpy suddenly stopped working, could be third parties. Probably needs some more digging, lets see what we can find out to make numpy ironclad to any dependency issue.

The solution is indeed related to some DLL depencency issues under WIN with Intel MKL libraries the numpy import error was under python 3.6.2 and numpy 1.13.1 :

"import numpy" numpy\core_init_.py", line 16, in from . import multiarray ImportError: DLL load failed: The specified procedure could not be found.

The quick and dirty solution is to copy all Intel mkl*.dll libraries from the specific environments (env) subfolder (seen at stackoverflow)

<\Library\bin> to <\Lib\site-packages\numpy\core>

04/13/2017  12:50 AM        30,277,392 mkl_avx.dll
04/13/2017  12:50 AM        38,253,840 mkl_avx2.dll
04/13/2017  12:50 AM        42,234,640 mkl_avx512.dll
04/13/2017  12:50 AM        45,722,896 mkl_avx512_mic.dll
04/13/2017  12:50 AM        25,369,360 mkl_core.dll
04/13/2017  12:50 AM        24,541,456 mkl_def.dll
04/13/2017  12:50 AM        24,506,640 mkl_intel_thread.dll
04/13/2017  12:50 AM        29,744,400 mkl_mc.dll
04/13/2017  12:50 AM        30,594,320 mkl_mc3.dll
04/13/2017  12:50 AM        12,567,824 mkl_rt.dll
04/13/2017  12:50 AM        11,388,688 mkl_sequential.dll
04/13/2017  12:50 AM        12,575,504 mkl_tbb_thread.dll
04/13/2017  12:50 AM        11,250,960 mkl_vml_avx.dll
04/13/2017  12:50 AM        11,463,952 mkl_vml_avx2.dll
04/13/2017  12:50 AM        10,285,840 mkl_vml_avx512.dll
04/13/2017  12:50 AM        11,692,304 mkl_vml_avx512_mic.dll
04/13/2017  12:50 AM         5,224,208 mkl_vml_cmpt.dll
04/13/2017  12:50 AM         5,345,552 mkl_vml_def.dll
04/13/2017  12:50 AM         9,477,904 mkl_vml_mc.dll
04/13/2017  12:50 AM         9,368,336 mkl_vml_mc2.dll
04/13/2017  12:50 AM         9,593,616 mkl_vml_mc3.dll

now numpy can be imported and runs fine. One last thing, the best approach is to match compatible versions of Python and numpy, which of course change as releases are made. See [numpy compatibility index] (https://numpy.org/neps/nep-0029-deprecation_policy.html) for up to date compatibility information. Don't forget to match scipy and matplotlib versions to numpy as well.

Answer 2

I fixed this problem with the following steps:

1. Creating and activating a new conda environment
2. Installing numpy within the environment
3. Then using this environment for my project

I was using PyCharm so to execute (3), I just set the project interpreter to the Python interpreter in the new environment I created in (1).

Answer 3

In my case this issue was related to numpy. So, I removed the numpy and again installed it. To remove numpy if you already have conda environment can we done by conda remove numpy if this command is removing some other packages also which are dependent on numpy then in the same environment if pip is installed, numpy can be removed using pip pip uninstall numpy and to again download it use pip install numpy. This should probably fix your error.

Answer 4

I seem to have gotten numpy installed for Python 2.7.13 on the Windows 10 64-bit computer using the following steps. I was able to install numpy for either a 32-bit or a 64-bit version of Python 2.7.13. Instructions for both are provided.

At the top of this answer are my instructions for 32-bit Python.
In the middle are my instructions for 64-bit Python. At the bottom are my instructions for 64-bit Python using a wheel file.

I suspect the primary problem in my original post may have arisen from not realizing that apparently the default version of Python that is downloaded from here (https://www.python.org/downloads/) is 32-bit.

For installing numpy on 32-bit Python 2.7.13

1. uninstalled Anaconda

2. removed Anaconda from User Variables in the Environment Variables window

3. installed Python 2.7.13 from the official Python website: https://www.python.org/

4. in command window typed:

   C:\users\general1>cd:\python27

5. Attempted to update pip in command window although pip was already up to date:

   python -m pip install -U pip

6. in command window used the following command, which returned a huge number of errors:

   pip install numpy

7. in command window typed:

   python -m pip install numpy

Step 7 seems to have worked. It returned the following message:

C:\Python27>python -m pip install numpy
Collecting numpy
  Retrying (Retry(total=4, connect=None, read=None, redirect=None)) 
  after connection broken by 'ProtocolError('Connection aborted.',
  error(10054, 'An existing connection was forcibly 
  closed by the remote host'))': /simple/numpy/
  Downloading numpy-1.12.0-cp27-none-win32.whl (6.6MB)
    100% |################################| 6.6MB 64kB/s
Installing collected packages: numpy
Successfully installed numpy-1.12.0

8. Opened Python 2.7.13 and typed:

   import numpy

Only the command prompt was returned. No errors or warnings were issued.

I have not yet tried to use any code that requires numpy, but at least it seems to have been installed.

Step 7 says that a 32-bit file was downloaded:

numpy-1.12.0-cp27-none-win32.whl

I do not know whether a 32-bit version of numpy was installed or a 64-bit version. Perhaps the file is a 64-bit version that was renamed to say it was a 32-bit version as described in my question.

I typed in Python (from Do I have Numpy 32 bit or 64 bit?):

import numpy.distutils.system_info as sysinfo
sysinfo.platform_bits
32

So, it seems a 32-bit version of numpy was installed although my Python is supposed to be a 64-bit version (from here: How do I determine if my python shell is executing in 32bit or 64bit mode on OS X?):

import platform
platform.architecture()[0]
'32bit'

Perhaps in all of the installing and uninstalling of Python I accidentally installed a 32-bit version the last time. I can try installing Python again and use the intended 64-bit version this time. Hopefully the same steps above will still work.

Perhaps if I want a 64-bit version of Python 2.7.13 I am to click on the following on the official Python webpage (here: https://www.python.org/downloads/windows/):

Download Windows x86-64 MSI installer

instead of on the prominent (https://www.python.org/downloads/):

Download Python 2.7.13

I am also curious to know from where the file numpy-1.12.0-cp27-none-win32.whl was downloaded. I will provide that too if I find out. Maybe it was downloaded from here: https://pypi.python.org/pypi/numpy

For installing numpy on 64-bit Python 2.7.13

The steps above worked for installing numpy for what turned out to be a 32-bit version of Python 2.7.13.

To install numpy for a 64-bit version of Python 2.7.13 I used the following steps:

1. uninstall Python 2.7.13

2. Download: python-2.7.13.amd64.msi by clicking on: Download Windows x86-64 MSI installer from here: https://www.python.org/downloads/windows/

3. in command window typed:

   C:\users\general1>cd:\python27

4. Attempted to update pip in command window although pip was already up to date:

   python -m pip install -U pip

5. in command window used the following command, which this time did not return any errors:

   pip install numpy

which returned:

C:\Python27>pip install numpy
Collecting numpy
  Downloading numpy-1.12.0-cp27-none-win_amd64.whl (7.5MB)
    100% |################################| 7.5MB 63kB/s
Installing collected packages: numpy
Successfully installed numpy-1.12.0

6. Opened Python 2.7.13 and typed:

   import numpy

No errors or warnings were returned.

7. Checked whether numpy was 64-bit or 32-bit. It was 64-bit.

   import numpy.distutils.system_info as sysinfo

   sysinfo.platform_bits

   64

For installing numpy wheel files in Python 2.7.13

After installing numpy I tried to install scipy and received an error: numpy.distutils.system_info.NotFoundError: no lapack/blas resources found (described here Windows Scipy Install: No Lapack/Blas Resources Found).

I did the following:

1. Uninstalled Python 2.7.13

2. Installed Visual Studio Community 2017 RC as described by drewid (here: Windows Scipy Install: No Lapack/Blas Resources Found)

3. Installed VCForPython27.msi also as described by drewid.

Then I received a scipy error cannot import name NUMPY_MKL mentioned here (ImportError: cannot import name NUMPY_MKL).

I downloaded the files numpy-1.12.0+mkl-cp27-cp27m-win_amd64.whl and scipy-0.18.1-cp27-cp27m-win_amd64.whl from here: http://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy as suggested by VMAtm.

Then I:

1. uninstall Python 2.7.13

2. Reinstall Python 2.7.13 using python-2.7.13.amd64.msi

3. Installed numpy-1.12.0+mkl-cp27-cp27m-win_amd64.whl from http://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy using the following line in the command window:

   pip install numpy-1.12.0+mkl-cp27-cp27m-win_amd64.whl

4. Installed scipy-0.18.1-cp27-cp27m-win_amd64.whl from http://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy using the following line in the command window

   pip install scipy-0.18.1-cp27-cp27m-win_amd64.whl

Both numpy and scipy can now import inside Python 2.7.13

Note that I placed scipy-0.18.1-cp27-cp27m-win_amd64.whl and numpy-1.12.0+mkl-cp27-cp27m-win_amd64.whl in the folder c:\Python27 prior to using the aforementioned pip commands.

C:\>cd\Python27

C:\Python27>pip install numpy-1.12.0+mkl-cp27-cp27m-win_amd64.whl
Processing c:\python27\numpy-1.12.0+mkl-cp27-cp27m-win_amd64.whl
Installing collected packages: numpy
Successfully installed numpy-1.12.0+mkl

C:\Python27>pip install scipy-0.18.1-cp27-cp27m-win_amd64.whl
Processing c:\python27\scipy-0.18.1-cp27-cp27m-win_amd64.whl
Installing collected packages: scipy
Successfully installed scipy-0.18.1

C:\Python27>