CODEWITHSUNDEEP
pythonnumpyscipyintel-mkl
Erik Iverson
Erik Iverson

Reputation: 353

change python distribution

I've run into problems import scipy.interpolate module. It uses numpy. I am on windows 7 and have managed python and all the modules without a distribution manager like anaconda. Pip is easy to use so that's how I do it. But part of numpy on windows can, and maybe needs to, leverage intel math kernel library. My problem is along the lines of Difference between Numpy and Numpy-MKL?

I've tried solutions described like this: How to install numpy+mkl for python 2.7 on windows 64 bit? Since I use python 3, and my processor is a Intel(R) Core(TM) i7-5600U CPU @ 2.60GHz, I've tried a few of the whl's on here: https://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy . I can't seem to pick the right one because pip will say that whl is not supported on my platform.

So should I just switch to the Intel Distribution for Python? https://software.intel.com/en-us/distribution-for-python. Will switching to a distribution mess up my python directory structures?

I also found this solution: NumPy 64bit fail to install with pip on Windows with Python 64bit but I'm not sure if editing pip itself should be necessary. Thanks.

Upvotes: 1

Views: 309

Answers (2)

Erik Iverson
Erik Iverson

Reputation: 353

Turns out I have 32 bit python, not 64. Choosing the latest (1.15) numpy enabled it to install, and now scipy.interpolate loads and runs.

Upvotes: 0

chjortlund
chjortlund

Reputation: 4017

I had the same problem some time ago, and what worked for me in the end was to uninstall numpy and scipy, download the correct 64 bit wheel from gohlke's webside and then rename the *win64.whl file to *win32.whl

For some reason when installing Numpy with MKL, pip was unable to detect the correct architecture.

Before doing this, I will recommend you to simply try to upgrade pip to the latest version (10.1), and try to install normally (this issue report looks related: https://bugs.python.org/issue18987).

It's also important to install Numpy+MKL before Scipy.

Hope it helps.

Upvotes: 2

Related Questions