0

Is there an option to install SciPy on Windows without installing Anaconda as well? I could not do it via pip and everywhere it says to use Anaconda.

More details:

  • I want the SciPy package without any additional programs like Python(x, y) or Canopy.
  • The error with pip is: numpy.distutils.system_info.NotFoundError: no lapack/blas resources found. From research I found that I need to use additional packages but it sounds strange to me. I couldn't install LAPACK or BLAS.
smci
  • 32,567
  • 20
  • 113
  • 146
MyNick
  • 536
  • 1
  • 9
  • 25
  • Well there are more scientific Python distributions besides Anaconda, like Canopy and Python (x,y) available. It can be done standalone as well. What platform are you attempting to install to? What went wrong with the pip install? It's hard to help without those details. You should edit the question to add them. – paisanco Sep 26 '15 at 16:07
  • This question may help although its a bit old: http://stackoverflow.com/questions/2213551/installing-scipy-with-pip – paisanco Sep 26 '15 at 16:15

2 Answers2

2

There are unofficial builds: http://www.lfd.uci.edu/~gohlke/pythonlibs. Here's a link to scipy: http://www.lfd.uci.edu/~gohlke/pythonlibs#scipy

Adobe
  • 12,967
  • 10
  • 85
  • 126
  • how to i install it? whl file – MyNick Sep 26 '15 at 16:27
  • scipy-0.16.0-cp27-none-win_amd64.whl is not a supported wheel on this platform. – MyNick Sep 26 '15 at 16:33
  • Worked! Thank you so much. By the way, do i have to download it and pass pip the path or i can give it the link somehow? – MyNick Sep 26 '15 at 16:42
  • Wait, i cant import doccer, it written in: http://stackoverflow.com/questions/24596594/cant-import-scipys-io that i need the 64 bit version (i got 64 bit python), what do i do? – MyNick Sep 26 '15 at 16:46
  • If you have x64 python, then you should be using scipy-0.16.0-cp27-none-win_amd64.whl. Let's try to investigate why it doesn't work. Can you run `python -c "import pip; print(pip.pep425tags.get_supported())"` from command line? (or just `import pip; print(pip.pep425tags.get_supported())` from python) – Adobe Sep 26 '15 at 16:56
  • [('cp27', 'none', 'win32'), ('py2', 'none', 'win32'), ('cp27', 'none', 'any'), ('cp2', 'none', 'any'), ('cp26', 'none', 'any'), ('cp25', 'none', 'any'), ('cp24', 'none', 'any'), ('cp23', 'none', 'any'), ('cp22', 'none', 'any'), ('cp21', 'none', 'any'), ('cp20', 'none', 'any'), ('py27', 'none', 'any'), ('py2', 'none', 'any'), ('py26', 'none', 'any'), ('py25', 'none', 'any'), ('py24', 'none', 'any'), ('py23', 'none', 'any'), ('py22', 'none', 'any'), ('py21', 'none', 'any'), ('py20', 'none', 'any')] Thank you for your time – MyNick Sep 26 '15 at 16:57
  • Seems like `'cp27', 'none', 'any'` means `scipy-0.16.0-cp27-none-win_amd64.whl` should work for you. – Adobe Sep 26 '15 at 17:00
0

You can proceed with installing numpy or scipy using pip, after you install lapack and blas, which are system libraries. It shouldn't be very hard, but depends on you OS.

For RedHat/CentOS/Fedora this could be done with:

yum install lapack lapack-devel blas blas-devel

The packages can be found e.g. in CentOS base repository.

However, the scikit-learn website says as follows:

We don’t recommend installing scipy or numpy using pip on linux, as this will involve a lengthy build-process with many dependencies. Without careful configuration, building numpy yourself can lead to an installation that is much slower than it should be. If you are using Linux, consider using your package manager to install scikit-learn. It is usually the easiest way, but might not provide the newest version. If you haven’t already installed numpy and scipy and can’t install them via your operation system, it is recommended to use a third party distribution.

Package managers are usually yum or apt-get and again on RedHat/CentOS/Fedora you can skip using pip and install this way:

yum install scipy

Third party distributions mentioned above are things like anaconda or Python(x,y).

Community
  • 1
  • 1
Dominik
  • 390
  • 3
  • 16
  • But the OP said they're on Windows. This is all much more painful on Windows. That's precisely why almost all Windows installs of Python with numeric libraries are distributions, not locally built. – smci Nov 17 '18 at 02:04