2

Task I want to install Scipy on 64bit python.

Already Done I have tried following ways:

  • Using pip install numpy/scipy-- in this case numpy installs well but scipy fails. error: can't find lpack..
  • Install Scipy from here -- in this it installs well but on importing gives an error not valid win32 application. As its 32bit and python is 64bit
  • Use gohlke to install scipy. Installation goes well but when importing any sub-module like import scipy.stats throws an error: specified module can not be found.

Reason to move on 64bit Python I have 40000 data points csv file. On ruining linkage function python throws a memory error. Some posts suggested to move to 64bit can solve the issue.

Improve this question
1
  • 1
    With all due respect to the Task Definition, let me comment on [Reason]. There are more important issues than a static .CSV size pushing one to move into 64bit arena. My ML/AI processing works smooth on 200.000+ x 100+ ( in float32 ( not because a 32bit O/S ) due to numpy .dtype set this way in the ML/AI module components). Yes, 64bit O/S will give you less worries about some issues, nevertheless python memory management is not forgiving once working with large dataSets/dataStructures, both in 32bit & the same in 64bit. Travis' Anaconda, as @rth noted, is tool of choice Commented Jul 13, 2015 at 11:30

2 Answers 2

Reset to default
1

Installing scientific Python modules from sources on Windows is a bit complex and not very maintainable.

A more reliable solution is to use a scientific python distribution, such as Anaconda or Enthought Canopy.

Edit: I understand this answer is opinion based, however I am not aware of other simple and open-source way to deploy scientific python modules on Windows. Sure you could install scipy from sources or with pip, however in a wider picture, if you need Numpy, Scipy, IPython and Matplotlib, etc. that's just not a workable solution.

Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

I am pretty sure that the entire community, developers included, agree with you on that compling from source is not realistic. Anaconda and EPD are great; I do think WinPython works even better if you use Cython, Numba and theano. Python(x,y) used to be good, but they seem to be behind on the 64bit versions.
1

Another clean solution is to download the wheel file for SciPy(compatiable with your architecture of Python) and use pip install pythonModule.whl and it should install SciPy with no problems. I had to do this the other day and I stumbled upon this website, which has many Python modules already compiled to binaries. Just another way to do it if you don't want to download Anaconda.

I did this for NumPy, SciPy, and Scikit-Learn.

Improve this answer

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.