Quantcast

Numpy Not Fully Installing

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Numpy Not Fully Installing

Gary Gaulin
I am having a problem installing numpy on my Windows 7 PC. The install seems to begin OK then Python reports:

building library "npymath" sources
No module named msvccompiler in numpy.distutils; trying from distutils
error: Unable to find vcvarsall.bat


No C compiler is installed, and I'm not sure which to use where I needed one. CUDA for my NVIDIA GeForce GT 520 might also no longer be installed, but I'm not sure whether it's best to install/update CUDA after getting Python set to go.

The steps I took to at least get bidict properly installed are as follows:

Downloaded Python from http://www.python.org/ and (automatically) installed it at C:\Python27\

Into a Command Prompt window (DOS shell) type: "set path=%path%;C:\Python27\"

To see the Registry file just changed use the Registry Editor: enter "RegEdit" into run (or find) textbox in "Start" menu.

Make a folder in Python27 named "Modules" for download & unzipping to a new "C:\Python27\Modules\" path for storing all in one easy to find place.

The following properly installed "simplejson".
http://www.wikihow.com/Install-Python-Packages-on-Windows-7

Save get-pip.py to a ""C:\Python27\Modules\Pip" folder then from Command Prompt window (from that directory) run "python get-pip.py"

To test pip and first install use: http://stackoverflow.com/questions/739993/how-can-i-get-a-list-of-locally-installed-python-modules

import pip
installed_packages = pip.get_installed_distributions()
installed_packages_list = sorted(["%s==%s" % (i.key, i.version)
for i in installed_packages])
print(installed_packages_list)


Output text shows example module was properly installed:

>>> ================================ RESTART ================================
>>>
['simplejson==3.6.4']
>>> _

To work with Command line Prompt: left click "Start", right click "Computer", left click "Properties", "Advanced system settings", "Environmental Variables". In the "System variables" list left click the variable named "Path" then the "Edit" button. Add to the end of the line ";C:\Python27\;C:\Python27\Scripts\" then close menus by clicking "OK" to all.
Test from Command Prompt by entering "python". If successful (which it was) then help information is listed and command prompt changes to three greater than symbols ">>>".
After making change Windows 7 PC may make a guitar type error sound following startup music?

Other modules to install for the FlyBrain project are at:
https://github.com/neurokernel/neurokernel/blob/master/INSTALL.rst#installation-dependencies

This is what I downloaded to install from. In one by one going through the list I downloaded the following, where some had .exe I saved in case the .gz does not work:  

bidict 0.1.0 or later. (already installed OK)
 bidict-0.2.1.tar.gz

bottleneck 0.7.0 or later.
 Bottleneck-0.8.0.tar.gz

futures 2.1.5 or later.
 futures-2.2.0.tar.gz

h5py 2.2.1 or later.
 h5py-2.3.1.tar.gz
 or
 h5py-2.3.1.win32-py2.7.exe

matplotlib 1.3.0 or later.
 matplotlib-1.4.0.tar.gz
 or
 matplotlib-1.4.0.win-amd64-py2.7.exe
 
msgpack-numpy 0.3.1.1 or later.
 msgpack-numpy-0.3.2.tar.gz

networkx 1.8 or later
 networkx-1.9.1.zip

numexpr 2.3 or later.
 numexpr-master.zip

numpy 1.2.0 or later.
 numpy-1.9.0.zip

pycuda 2012.1 or later.
 pycuda-2014.1.tar.gz

pyzmq 13.0 or later.
 pyzmq-14.3.1.zip

scipy 0.11.0 or later.
 scipy-master.zip

tables 2.4.0 or later.
 tables-3.1.1.tar.gz

twiggy 0.4.0 or later.
 twiggy-maint-0.4.zip

Also but not installed is one that I'm not sure whether it's needed or not:
 virtualenv-1.11.6.tar.gz

This is my first experience with such a complex Python installation. Your Fly Brain software makes it worth updating from Visual Basic 6.0 that still works fine for programming minimal code models, but not detailed neuron by neuron models of insect brains like you are putting together.
http://intelligencegenerator.blogspot.com/

All in all things still went well for finding all the downloads. I hope my experience is a help for you, and for others who may make the same beginners mistakes that I seem to be making.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Numpy Not Fully Installing

Lev Givon
Administrator
Received from Gary Gaulin on Wed, Oct 01, 2014 at 01:25:20AM EDT:
> I am having a problem installing numpy on my Windows 7 PC. The install seems
> to begin OK then Python reports:
>
> /building library "npymath" sources
> No module named msvccompiler in numpy.distutils; trying from distutils
> error: Unable to find vcvarsall.bat /
>
> No C compiler is installed, and I'm not sure which to use where I needed
> one.

There are prebuilt numpy packages available for Windows:
http://www.scipy.org/install.html#individual-binary-and-source-packages

> CUDA for my NVIDIA GeForce GT 520 might also no longer be installed,
> but I'm not sure whether it's best to install/update CUDA after getting
> Python set to go.

You do need to have CUDA installed before attempting to build PyCUDA.

> The steps I took to at least get bidict properly installed are as follows:

(snip)

> To test pip and first install use:
> http://stackoverflow.com/questions/739993/how-can-i-get-a-list-of-locally-installed-python-modules
>
> /import pip
> installed_packages = pip.get_installed_distributions()
> installed_packages_list = sorted(["%s==%s" % (i.key, i.version)
> for i in installed_packages])
> print(installed_packages_list)/

You can also use https://pypi.python.org/pypi/yolk3k
to list packages from the command line.

(snip)

> Also but not installed is one that I'm not sure whether it's needed or not:
>  virtualenv-1.11.6.tar.gz
>
> This is my first experience with such a complex Python installation. Your
> Fly Brain software makes it worth updating from Visual Basic 6.0 that still

If you have pip installed (with or without virtualenv) it should be able to
install most of the packages above in one fell swoop without you having to manually
download anything. Several packages do need to be installed before the others
owing to limitations in how the latter packages' respective  installation
procedures handle their Python dependencies

> works fine for programming minimal code models, but not detailed neuron by
> neuron models of insect brains like you are putting together.
> http://intelligencegenerator.blogspot.com/
>
> All in all things still went well for finding all the downloads. I hope my
> experience is a help for you, and for others who may make the same beginners
> mistakes that I seem to be making.
--
Lev Givon
Bionet Group | Neurokernel Project
http://www.columbia.edu/~lev/
http://lebedov.github.io/
http://neurokernel.github.io/

_______________________________________________
neurokernel-dev mailing list
[hidden email]
https://lists.columbia.edu/mailman/listinfo/neurokernel-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Numpy Not Fully Installing

Daniel Salles Chevitarese
I think this website is a very good choice: http://www.lfd.uci.edu/~gohlke/pythonlibs/

They also have packages compiled against Intel MKL, which gives a boost in some cases.

Daniel Salles Chevitarese
http://daniel.chevitarese.com.br

2014-10-01 9:47 GMT-03:00 Lev Givon <[hidden email]>:
Received from Gary Gaulin on Wed, Oct 01, 2014 at 01:25:20AM EDT:
> I am having a problem installing numpy on my Windows 7 PC. The install seems
> to begin OK then Python reports:
>
> /building library "npymath" sources
> No module named msvccompiler in numpy.distutils; trying from distutils
> error: Unable to find vcvarsall.bat /
>
> No C compiler is installed, and I'm not sure which to use where I needed
> one.

There are prebuilt numpy packages available for Windows:
http://www.scipy.org/install.html#individual-binary-and-source-packages

> CUDA for my NVIDIA GeForce GT 520 might also no longer be installed,
> but I'm not sure whether it's best to install/update CUDA after getting
> Python set to go.

You do need to have CUDA installed before attempting to build PyCUDA.

> The steps I took to at least get bidict properly installed are as follows:

(snip)

> To test pip and first install use:
> http://stackoverflow.com/questions/739993/how-can-i-get-a-list-of-locally-installed-python-modules
>
> /import pip
> installed_packages = pip.get_installed_distributions()
> installed_packages_list = sorted(["%s==%s" % (i.key, i.version)
> for i in installed_packages])
> print(installed_packages_list)/

You can also use https://pypi.python.org/pypi/yolk3k
to list packages from the command line.

(snip)

> Also but not installed is one that I'm not sure whether it's needed or not:
>  virtualenv-1.11.6.tar.gz
>
> This is my first experience with such a complex Python installation. Your
> Fly Brain software makes it worth updating from Visual Basic 6.0 that still

If you have pip installed (with or without virtualenv) it should be able to
install most of the packages above in one fell swoop without you having to manually
download anything. Several packages do need to be installed before the others
owing to limitations in how the latter packages' respective  installation
procedures handle their Python dependencies

> works fine for programming minimal code models, but not detailed neuron by
> neuron models of insect brains like you are putting together.
> http://intelligencegenerator.blogspot.com/
>
> All in all things still went well for finding all the downloads. I hope my
> experience is a help for you, and for others who may make the same beginners
> mistakes that I seem to be making.
--
Lev Givon
Bionet Group | Neurokernel Project
http://www.columbia.edu/~lev/
http://lebedov.github.io/
http://neurokernel.github.io/

_______________________________________________
neurokernel-dev mailing list
[hidden email]
https://lists.columbia.edu/mailman/listinfo/neurokernel-dev


_______________________________________________
neurokernel-dev mailing list
[hidden email]
https://lists.columbia.edu/mailman/listinfo/neurokernel-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Numpy Not Fully Installing

Gary Gaulin
In reply to this post by Gary Gaulin
That worked Lev! And thanks Daniel for the other help that answered several more of my questions, especially this:
"Many binaries depend on Numpy-MKL 1.8 and the Microsoft Visual C++ 2008 (x64, x86, and SP1 for CPython 2.6 to 3.2) or Visual C++ 2010 (x64, x86, for CPython 3.3 and 3.4) redistributable packages."

I started fresh by uninstalling all Python related software and changes to path. Anaconda-2.1.0-Windows-x86_64.exe was downloaded from http://continuum.io/downloads

I was not sure whether to choose the 32 bit version and went with the main choice on the page, the 64.

Clicking on the file installed the whole thing. From the Command Prompt I ran the pip test program (with added code to also save the list to a disk file to that directory) using the command "python modulestxt.py"

# File: ModulesTxt.py
# Save a list of all installed modules to disk file named Modules.txt
f = open('Modules.txt', 'w')
print f
# From http://stackoverflow.com/questions/739993/how-can-i-get-a-list-of-locally-installed-python-modules
import pip
installed_packages = pip.get_installed_distributions()
installed_packages_list = sorted(["%s==%s" % (i.key, i.version)
     for i in installed_packages])
print(installed_packages_list)
s = str(installed_packages_list)
f.write(s)
f.close()

A copy/paste from the Modules.txt file shows I now have all this installed:

['argcomplete==0.8.1', 'astropy==0.4.2', 'atom==0.3.9', 'backports.ssl-match-hostname==3.4.0.2', 'beautifulsoup4==4.3.2', 'binstar==0.7.1', 'bitarray==0.8.1', 'blaze==0.6.3', 'blz==0.6.2', 'bokeh==0.6.1', 'boto==2.32.1', 'casuarius==1.1', 'cdecimal==2.3', 'cffi==0.8.6', 'chaco==4.4.1', 'colorama==0.3.1', 'conda-build==1.8.2', 'conda==3.7.0', 'configobj==5.0.6', 'cryptography==0.5.4', 'cython==0.21', 'cytoolz==0.7.0', 'datashape==0.3.0', 'decorator==3.4.0', 'docutils==0.12', 'enable==4.3.0', 'enaml==0.9.8', 'flask==0.10.1', 'future==0.13.1', 'futures==2.1.6', 'gevent-websocket==0.9.3', 'gevent==1.0.1', 'greenlet==0.4.4', 'grin==1.2.1', 'h5py==2.3.1', 'ipython==2.2.0', 'itsdangerous==0.24', 'jdcal==1.0', 'jinja2==2.7.3', 'kiwisolver==0.1.3', 'llvmpy==0.12.7', 'lxml==3.4.0', 'markupsafe==0.23', 'matplotlib==1.4.0', 'menuinst==1.0.4', 'mock==1.0.1', 'multipledispatch==0.4.7', 'networkx==1.9.1', 'nltk==3.0.0', 'nose==1.3.4', 'numba==0.14.0', 'numexpr==2.3.1', 'numpy==1.9.0', 'openpyxl==1.8.5', 'pandas==0.14.1', 'patsy==0.3.0', 'pep8==1.5.7', 'pil==1.1.7', 'ply==3.4', 'psutil==2.1.1', 'py==1.4.25', 'pycosat==0.6.1', 'pycparser==2.10', 'pycrypto==2.6.1', 'pyface==4.4.0', 'pyflakes==0.8.1', 'pygments==1.6', 'pyopenssl==0.14', 'pyparsing==2.0.1', 'pyreadline==2.0', 'pytest==2.6.3', 'python-dateutil==1.5', 'pytz==2014.7', 'pywin32==219', 'pyyaml==3.11', 'pyzmq==14.3.1', 'requests==2.4.1', 'rope==0.9.4', 'runipy==0.1.1', 'scikit-image==0.10.1', 'scikit-learn==0.15.2', 'scipy==0.14.0', 'six==1.8.0', 'sockjs-tornado==1.0.1', 'sphinx==1.2.3', 'spyder==2.3.1', 'sqlalchemy==0.9.7', 'statsmodels==0.5.0', 'sympy==0.7.5', 'tables==3.1.1', 'toolz==0.7.0', 'tornado==4.0.2', 'traits==4.4.0', 'traitsui==4.4.0', 'ujson==1.33', 'unicodecsv==0.9.4', 'werkzeug==0.9.6', 'wsgiref==0.1.2', 'xlrd==0.9.3', 'xlsxwriter==0.5.7', 'xlwings==0.2.2', 'xlwt==0.7.5']

Numpy (and others) are clearly in the list.

I complicated things by trying to make it as simple as possible to see what is going in for modules and where they start up from on my hard drive. After seeing all I was in for that way your answer and the Scipy package probably saved me weeks of work. For the sake of making it easy for others I was hoping that the install would be as simple as possible and so far it turned out easier than it at first seemed.

I now have CUDA to reinstall, and All Programs from Start button has a "Anaconda (64-bit)" menu of other things to look through. At this point I have to consider my problem getting numpy installed to have been well resolved, and the rest is looking easier. But I'll make sure to let you know if I then somehow manage to get stuck along the way again, by overcomplicating things for myself or whatever.
Loading...