{ "id": 12677, "title": "Install Theano on Windows 8.1 with GPU enabled: pycuda installation problems", "url": "/competitions/datasciencebowl/discussion/12677", "author_name": "Buzz Lightyear", "post_date": "2015-03-04T11:03:15.457000", "votes": 0, "comment_count": 29, "views": 29331, "content": "

Dear All,

\n

I'm getting crazy on installing Theano. I've a laptop with two graphic cards, one of them being NVIDIA GeForce GT 750M and Windows 8.1 as operating system.

\n

I followed the instruction on 

\n

http://stackoverflow.com/questions/25729969/installing-theano-on-windows-8-with-gpu-enabled

\n
    \n
  1. Microsoft Visual Studio 12.0 installation;
    2. \n
  2. CUDA 6.5 installation (64-bit Notebook version). It was successful (I launched deviceQuery);
    3. \n
  3. Python, Git, MinnGW, Cygwin installation;
    4. \n
  4. Modified the msvc9compiler.py file and added the line: ld_args.append('/MANIFEST')
    5. \n
  5. PyCUDA installation doesn't work!
    6. \n
\n\n\n

BOOST_INC_DIR = []
BOOST_LIB_DIR = []
BOOST_COMPILER = 'gcc43'
USE_SHIPPED_BOOST = True
BOOST_PYTHON_LIBNAME = ['boost_python-py27']
BOOST_THREAD_LIBNAME = ['boost_thread']
CUDA_TRACE = False
CUDA_ROOT = 'C:\\\\Program Files\\\\NVIDIA GPU Computing Toolkit\\\\CUDA\\\\v6.5'
CUDA_ENABLE_GL = False
CUDA_ENABLE_CURAND = True
CUDADRV_LIB_DIR = ['${CUDA_ROOT}/lib', '${CUDA_ROOT}/lib/x64']
CUDADRV_LIBNAME = ['cuda']
CUDART_LIB_DIR = ['${CUDA_ROOT}/lib', '${CUDA_ROOT}/lib/x64']
CUDART_LIBNAME = ['cudart']
CURAND_LIB_DIR = ['${CUDA_ROOT}/lib', '${CUDA_ROOT}/lib/x64']
CURAND_LIBNAME = ['curand']
CXXFLAGS = ['/EHsc']
LDFLAGS = ['/FORCE']

\n\n

(my cuda path is C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5\\lib\\x64)

\n

6. then executed the following commands: 

\n

set VS90COMNTOOLS=%VS100COMNTOOLS%
python setup.py build
python setup.py install

\n

I don't receive any error.

\n

7. Created this python file:

\n

# from: http://documen.tician.de/pycuda/tutorial.html
import pycuda.gpuarray as gpuarray
import pycuda.driver as cuda
import pycuda.autoinit
import numpy
a_gpu = gpuarray.to_gpu(numpy.random.randn(4,4).astype(numpy.float32))
a_doubled = (2*a_gpu).get()
print a_doubled
print a_gpu

\n\n

and here there's something wrong (please see the screenshot attached).

\n

Would you give me a hand in understanding how to fix it?

\n

Thank you so much

\n

Cristian

", "messages": [ { "id": 65473, "postDate": "2015-03-05T04:44:30.150Z", "content": "

hi there, many of these cool toolkits have been developed on linux systems.  I recommend installing something like ubuntu and then install/build with only a couple of commands.

\n

I can understand one being apprehensive about committing to a dual boot...which does carry risks.  Grab a 3 year old desktop from the recycling and put a used gpu in it (gtx 580 for 100 bucks!) and that'll get you going. Using linux these days is a lot easier than it used to be.

\n

Once you do this, you will never go back!  Pinky-swear!

", "votes": 4 }, { "id": 77595, "postDate": "2015-05-08T10:17:25.193Z", "content": "

wait  a minute... The important error message here seems to be: 

\n

nvcc fatal : redefinition of argument 'machine'

\n

before reinstalling check it on google! 

\n

----------------------------------------------------------------------------------------------------------------------------------

\n\n

- pycuda is not necessary for theano working

\n

- I wrote microsoft visual studio 12 + cuda 6.5

\n

- to install theano:

\n", "votes": 1 }, { "id": 77574, "postDate": "2015-05-08T06:21:48.863Z", "content": "

Hello Mohammad,

\n

You get the error not because of that file missing, which indeed is in your file system.

\n

Try not to use that line (flags='-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include') in your .theanorc.txt) and check if it works.

\n

It's a real pain.... :-)

", "votes": 1 }, { "id": 65768, "postDate": "2015-03-09T10:34:13.543Z", "content": "

It was a pain, but at the end I've got Theano installed on a Windows 64-bit laptop.

\n

I had to uninstall and re-install Visual Studio, because of the damn crtdefs.h missing.

\n

Ciao!

\n

Cristian

", "votes": 1 }, { "id": 65533, "postDate": "2015-03-05T19:54:45.027Z", "content": "

Well, i was able to do gpu on windows (nolearn, lasagne and theano), but that was not easy.

\n

I used msys + mingw32 + win-builds (http://win-builds.org/1.5.0/msys-cygwin.html#_msys_installation)

\n

then installed cuda 6.5 & python2.7

\n

For theano, i used the prebuilds packages from here : http://www.lfd.uci.edu/~gohlke/pythonlibs/#pil

\n

I had to install pthread port : https://sourceware.org/pthreads-win32/

\n

I had to change some python files (didn't write it up and don't remember which files, but i think it was to configure properly pthread in theano ).

\n

And finally, there is a glitch with pylearn2 + cuda convnet.

\n

I had to add cuda_ndarray.pyd and libcuda_ndarray.pyd (got the two files from a tmp folder when python is compiling the cuda_convnet wrapper) to a folder in my PATH env variable.

\n

Hope this will help ! :)

\n\n

Edit : btw, here is my theanorc : 

\n

[global]
device = gpu1
floatX = float32
nocleanup = True

\n

[nvcc]
flags=-LC:\\Python27\\libs
compiler_bindir=C:\\Program Files (x86)\\Microsoft Visual Studio 10.0\\VC\\bin
fastmath = True

", "votes": 1 }, { "id": 65507, "postDate": "2015-03-05T13:29:34.737Z", "content": "

[quote=zero zero;65473]

\n

I can understand one being apprehensive about committing to a dual boot...which does carry risks. 

\n

[/quote]

\n

I finally gave up a few weeks ago trying to do GPU on windows. My solution was to install Ubuntu on a completely different and brand-new drive of my desktop. To switch environments, I just flip the boot priority when restarting.

\n

It's worked great.

", "votes": 2 }, { "id": 136538, "postDate": "2016-09-22T18:41:15.547Z", "content": "

http://web.archive.org/web/20150601112123/http://machinelearning.berlin/

", "rawMarkdown": "http://web.archive.org/web/20150601112123/http://machinelearning.berlin/" }, { "id": 135063, "postDate": "2016-09-10T17:00:21.877Z", "content": "

[quote=Christian Schröder de Witt;80419]

\n

Hi, try this tutorial here, it's about installing Theano on Windows 8.1 x64:

\n

http://machinelearning.berlin/?p=383

\n

I wrote it because I couldn't find anything that worked. If you stick exactly to the tutorial, it should work (you have few choices though).

\n

All the best Christian

\n

[/quote]

\n

Anyone knows if the tutorial is still availible somewhere else?

" }, { "id": 80419, "postDate": "2015-05-29T21:23:00.453Z", "content": "

Hi, try this tutorial here, it's about installing Theano on Windows 8.1 x64:

\n

http://machinelearning.berlin/?p=383

\n

I wrote it because I couldn't find anything that worked. If you stick exactly to the tutorial, it should work (you have few choices though).

\n

All the best Christian

" }, { "id": 77589, "postDate": "2015-05-08T09:42:21.273Z", "content": "

But first check one more time your theanorc file.

\n

Does it have this line?

\n

[cuda]

\n

root=your_path_to_cuda (v7.0)

\n

root=C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5 <--

" }, { "id": 77588, "postDate": "2015-05-08T09:37:14.700Z", "content": "

Not being an expert, I'd suggest to unistall cuda 7.0 and install cuda 6.5, which is a version which has been demonstrated to work good on my windows laptop :-) together with Microsoft Visual Studio 12.0.

\n

Happy fighting :-)

\n

Let me know about it!

" }, { "id": 77506, "postDate": "2015-05-07T18:42:36.893Z", "content": "

[quote=Buzz Lightyear;65768]

\n

It was a pain, but at the end I've got Theano installed on a Windows 64-bit laptop.

\n

I had to uninstall and re-install Visual Studio, because of the damn crtdefs.h missing.

\n

Ciao!

\n

Cristian

\n

[/quote]

Hello Buzz, i read this post for many times until i got to your last problem :

ValueError: Theano nvcc.flags support only parameter/value pairs without space between them. e.g.: '--machine 64' is not supported, but '--machine=64' is supported. Please add the '=' symbol. nvcc.flags value is ''-ID:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include''

So i check for the missing file crtdefs.h, but it was on my system actually! And i have still the last problem what you suggest

" }, { "id": 73189, "postDate": "2015-04-22T15:51:47.517Z", "content": "

Ciao Cristian,

\n

I did that already and placed the file in the proper dir, my problem is that I don't know how to replace that specific row with a row customized for Anaconda. Do you mean that in Anaconda that call is not needed so I can basically drop it or should I replace it? (therefore my question, I wouldn't know how)
As for now I'll try proceeding that way but I feel clumsier as I move on, lots of uncertainty, it's not my specialty I must admit.

\n

Thanks for you time,

\n

ciao,

\n

Luca

" }, { "id": 73173, "postDate": "2015-04-22T14:19:20.880Z", "content": "

Ciao Luca,

\n

That call is aimed at add winpython stuff. But you don't have to configure that file (it's located here C:\\SciSoft\\WinPython-64bit-2.7.9.3\\scripts\\env.bat on my laptop). You should configure an other env.bat (which is located here "C:\\Scisoft\\env.bat" on my laptop) as I've mentioned above.

\n

Fammi sapere

\n

Ciao

\n

Cristian

" }, { "id": 73144, "postDate": "2015-04-22T11:02:36.747Z", "content": "

Hi Buzz Lightyear,

\n

I’m trying to install Theano on my laptop as well, struggling quite a bit, maybe you can help. I’m using Anaconda and I’m following step by step the guide provided here: http://theano.readthedocs.org/en/latest/install_windows.html. The same process you have been following. I’m stuck with the environment configuration script 'env.bat' needed to configure the system path. The tutorial example refers to a WinPython distribution but as I’m installing in Anaconda and I don’t know how to configure that following row:

\n

CALL %SCISOFT%\\WinPython-64bit-2.7.9.4\\scripts\\env.bat

\n

 What location and file should I call in Anaconda (or since you are using WinPython) what is that Call aimed at in WinPython? so that I can try finding the same thingh in anaconda.

\n

Thanks,

\n

Saluti,

\n

Luca

" }, { "id": 65575, "postDate": "2015-03-06T08:06:10.910Z", "content": "

Hi Gregory,

\n

If I had that line:

\n

flags=-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include

\n

or

\n

flags='-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include'

\n

to theanorc I got an other error message:

\n

ValueError: Theano nvcc.flags support only parameter/value pairs without space between them. e.g.: '--machine 64' is not supported, but '--machine=64' is supported. Please add the '=' symbol. nvcc.flags value is '-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include'

\n

Any other help/hint?

\n

Thank you

\n

Cristian

" }, { "id": 65547, "postDate": "2015-03-05T21:23:14.790Z", "content": "

In your theanorc, i gess something like this should do the trick :

\n

[nvcc]

\n

flags=-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include

\n\n

Or you could add the visual include folder to your PATH (ugly)

" }, { "id": 65545, "postDate": "2015-03-05T21:10:07.777Z", "content": "

Hi Gregory, thank you for your help!

\n

Could you be even more clear please? :-)

\n

How do I add that path as include path?

" }, { "id": 65543, "postDate": "2015-03-05T20:54:04.723Z", "content": "

Yes, it can't find "crtdefs.h". Either add the following path as include path : "C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include" (-I option), either use vcvarsall to properly setup your command prompt.

" }, { "id": 65539, "postDate": "2015-03-05T20:27:22.230Z", "content": "

It's just a pain.... Now theano works but doesn't use the gpu. Arghhhhh!!!

\n

This is my laptop config is x64, nvcc/cuda 6.5, Python 2.7.9, WinPython-64bit-2.7.9.3, Windows 8.1, VS2013 and two graphic units (Intel HD Graphics 4600 and NVIDIA GeForce GT 750M).

\n

where gcc:
C:\\SciSoft\\TDM-GCC-64\\bin\\gcc.exe
C:\\Rtools\\gcc-4.6.3\\bin\\gcc.exe
C:\\MinGW\\bin\\gcc.exe

\n

where gendef:
C:\\SciSoft\\TDM-GCC-64\\x86_64-w64-mingw32\\bin\\gendef.exe

\n

where cl:
C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0\\VC\\bin\\amd64\\cl.exe
C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64\\cl.exe

\n

where nvcc:
C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5\\bin\\nvcc.exe

\n

I did a manual installation. Got the latest bleeding edge version GitHub and downloaded the latest zip. The .theanorc.txt file is located in C:\\SciSoft\\WinPython-64bit-2.7.9.3\\settings:

\n

[global]
device=gpu
floatX=float32

\n

[cuda]
root=C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5

\n

[nvcc]
fastmath=True
compiler_bindir=C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64
cxx=C:\\SciSoft\\TDM-GCC-64\\bin

\n

[blas]
ldflags=

\n

And here is what I get from a test....

\n

c:\\program files\\nvidia gpu computing toolkit\\cuda\\v6.5\\include\\host_config.h(149) : fatal error C1083: Impossibile aprire il file inclusione 'crtdefs.h': No such file or directory ERROR (theano.sandbox.cuda): Failed to compile cuda_ndarray.cu: ('nvcc return status', 2, 'for cmd', 'nvcc -shared -O3 --compiler-bindir C:\\\\Program Files (x86)\\\\Microsoft Visual Studio 12.0\\\\VC\\\\bin\\\\amd64 -Xlinker /DEBUG -D HAVE_ROUND -m64 -Xcompiler -DCUDA_NDARRAY_CUH=7772254ca6db84d79e28ecddbe019d53,-D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD -Ic:\\\\scisoft\\\\theano-master\\\\theano\\\\sandbox\\\\cuda -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\lib\\\\site-packages\\\\numpy\\\\core\\\\include -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\include -o C:\\\\Users\\\\Cris\\\\AppData\\\\Local\\\\Theano\\\\compiledir_Windows-8-6.2.9200-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.9-64\\\\cuda_ndarray\\\\cuda_ndarray.pyd mod.cu -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\libs -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64 -lpython27 -lcublas -lcudart') ERROR:theano.sandbox.cuda:Failed to compile cuda_ndarray.cu: ('nvcc return status', 2, 'for cmd', 'nvcc -shared -O3 --compiler-bindir C:\\\\Program Files (x86)\\\\Microsoft Visual Studio 12.0\\\\VC\\\\bin\\\\amd64 -Xlinker /DEBUG -D HAVE_ROUND -m64 -Xcompiler -DCUDA_NDARRAY_CUH=7772254ca6db84d79e28ecddbe019d53,-D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD -Ic:\\\\scisoft\\\\theano-master\\\\theano\\\\sandbox\\\\cuda -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\lib\\\\site-packages\\\\numpy\\\\core\\\\include -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\include -o C:\\\\Users\\\\Cris\\\\AppData\\\\Local\\\\Theano\\\\compiledir_Windows-8-6.2.9200-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.9-64\\\\cuda_ndarray\\\\cuda_ndarray.pyd mod.cu -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\libs -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64 -lpython27 -lcublas -lcudart') WARNING (theano.sandbox.cuda): CUDA is installed, but device gpu is not available (error: cuda unavilable) WARNING:theano.sandbox.cuda:CUDA is installed, but device gpu is not available (error: cuda unavilable)

\n

mod.cu

\n

['nvcc', '-shared', '-O3', '-use_fast_math', '--compiler-bindir', 'C:\\\\Program Files (x86)\\\\Microsoft Visual Studio 12.0\\\\VC\\\\bin\\\\amd64', '-Xlinker', '/DEBUG', '-D HAVE_ROUND', '-m64', '-Xcompiler', '-DCUDA_NDARRAY_CUH=7772254ca6db84d79e28ecddbe019d53,-D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD', '-Ic:\\\\scisoft\\\\theano-master\\\\theano\\\\sandbox\\\\cuda', '-IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\lib\\\\site-packages\\\\numpy\\\\core\\\\include', '-IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\include', '-o', 'C:\\\\Users\\\\Cris\\\\AppData\\\\Local\\\\Theano\\\\compiledir_Windows-8-6.2.9200-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.9-64\\\\cuda_ndarray\\\\cuda_ndarray.pyd', 'mod.cu', '-LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\libs', '-LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64', '-lpython27', '-lcublas', '-lcudart']
NP time: 0.116000[s], theano time: 0.114000[s] (times should be close when run on CPU!)
Result difference: 0.000000

\n

Does anybody know why it doesn't use the gpu? Cuda is installed and pass the deviceQuery test!

\n

I've been spending days on it....

\n

Thank you for your help

\n

Cristian

" }, { "id": 65538, "postDate": "2015-03-05T20:15:59.717Z", "content": "

I wasted countless hours/days trying to get this working on Windows 8.1. I finally gave up on this. It is just not very well supported. Best to stick with Linux distributions with and without GPU support.

" }, { "id": 65496, "postDate": "2015-03-05T12:06:17.977Z", "content": "

Hi,

\n

Here Theano is still not working....
This is my laptop config is x64, nvcc/cuda 6.5, Python 2.7.9, WinPython-64bit-2.7.9.3, Windows 8.1, VS2013 and two graphic units (Intel HD Graphics 4600 and NVIDIA GeForce GT 750M).

\n

This is my env.bat in c:\\SciSoft

\n

REM configuration of paths
set VSFORPYTHON="C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0"
set SCISOFT=%~dp0REM add tdm gcc stuff
set PATH=%SCISOFT%\\TDM-GCC-64\\bin;%SCISOFT%\\TDM-GCC-64\\x86_64-
w64-mingw32\\bin;%PATH%REM add winpython stuff
CALL %SCISOFT%\\WinPython-64bit-2.7.6.4\\scripts\\env.batREM configure path for msvc compilers
REM for a 32 bit installation change this line to
REM CALL %VSFORPYTHON%\\vcvarsall.bat
CALL %VSFORPYTHON%\\vcvarsall.bat amd64REM return a shell
cmd.exe /k
where gcc:
C:\\SciSoft\\TDM-GCC-64\\bin\\gcc.exe
C:\\Rtools\\gcc-4.6.3\\bin\\gcc.exe
C:\\MinGW\\bin\\gcc.exe

\n

where gendef:
C:\\SciSoft\\TDM-GCC-64\\x86_64-w64-mingw32\\bin\\gendef.exe

\n

where cl:
C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0\\VC\\bin\\amd64\\cl.exe
C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64\\cl.exe

\n

where nvcc:
C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5\\bin\\nvcc.exe

\n

I uninstalled and re-installed Theano from github following the instructions.
The .theanorc.txt file is located in C:\\SciSoft\\WinPython-64bit-2.7.9.3\\settings:

\n


[global]
device = gpu
floatX = float32

\n

[nvcc]

\n

compiler_bindir=C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64
cxx = C:\\SciSoft\\TDM-GCC-64\\bin
flags = --use-local-env --cl-version=2008

\n

But still not working.
Theano seems installed.... the main error message is "global name 'CVM' is not defined"

\n

Please check the notebook enclosed...
Thank you very much for your help

\n

Regards
Cristian

" }, { "id": 65421, "postDate": "2015-03-04T18:01:44.993Z", "content": "

I tried an other way, following http://deeplearning.net/software/theano/install_windows.html#install-windows.

\n

1- Visual Studio 2013 and Cuda 6.5 installed.

\n

2- Install Microsoft Visual C++ Compiler for Python 2.7

\n\n

3- Install TDM GCC in C:\\SciSoft\\TDM-GCC-64

\n

4- Install WinPython in C:\\SciSoft\\WinPython-64bit-2.7.9.3

\n

5- Configuring the environment. Saved the following shell script into c:\\scisoft\\env.bat to configure the system path:

\n

REM configuration of paths
set VSFORPYTHON="C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0"
set SCISOFT=%~dp0

REM add tdm gcc stuff
set PATH=%SCISOFT%\\TDM-GCC-64\\bin;%SCISOFT%\\TDM-GCC-64\\x86_64-w64-mingw32\\bin;%PATH%

REM add winpython stuff
CALL %SCISOFT%\\WinPython-64bit-2.7.9.3\\scripts\\env.bat

REM configure path for msvc compilers
REM for a 32 bit installation change this line to
REM CALL %VSFORPYTHON%\\vcvarsall.bat
CALL %VSFORPYTHON%\\vcvarsall.bat amd64

REM return a shell
cmd.exe /k

\n

6- create a link library for GCC. I opened the WinPython Command Prompt from C:\\SciSoft\\WinPython-64bit-2.7.9.3 and executed:

\n

gendef WinPython-64bit-2.7.9.3\\python-2.7.9.amd64\\python27.dll
dlltool --dllname python27.dll --def python27.def --output-lib WinPython-64bit-2.7.9.3\\python-2.7.9.amd64\\libs\\libpython27.a

\n


7- install Theano from github in C:\\SciSoft\\Theano

\n

8- from C:\\SciSoft\\Theano

\n

python setup.py develop

\n

9- configure Theano for GPU use (.theanorc.txt file) placed in C:\\scisoft\\WinPython-64bit-2.7.9.3\\settings

\n

[global]
device = gpu
floatX = float32

[nvcc]
flags = --use-local-env --cl-version=2008

\n

10- run a test http://deeplearning.net/software/theano/tutorial/using_gpu.html#using-gpu

\n

I got an Error Message... please check the attached file....

\n\n

Thank you so much for your help!

\n

I won't give up :-)

" }, { "id": 65410, "postDate": "2015-03-04T14:20:20.243Z", "content": "

@ sedielem: many thanks, but still in big troubles... As I wrote before cuda 6.5 is installed.

\n

Now I'm trying to install theano. I open Git Bash and I clone from github.

\n

git clone git://github.com/Theano/Theano.git

\n

Then from the folder where Theano files are:

\n

python setup.py install

\n

I got this long message error (please check the attached file).

\n

That's a nightmare...

\n

Thank you for your help!

" }, { "id": 65403, "postDate": "2015-03-04T12:03:45.117Z", "content": "

you do not need PyCUDA unless you plan to use the FFT-based convolution implementation. As far as I'm aware there are no other components in Theano right now that depend on PyCUDA.

\n

If you want to install it anyway, I'm afraid I can't help you with that - but it shouldn't be necessary :)

" }, { "id": 65396, "postDate": "2015-03-04T11:03:15.457Z", "content": "

Dear All,

\n

I'm getting crazy on installing Theano. I've a laptop with two graphic cards, one of them being NVIDIA GeForce GT 750M and Windows 8.1 as operating system.

\n

I followed the instruction on 

\n

http://stackoverflow.com/questions/25729969/installing-theano-on-windows-8-with-gpu-enabled

\n
    \n
  1. Microsoft Visual Studio 12.0 installation;
    2. \n
  2. CUDA 6.5 installation (64-bit Notebook version). It was successful (I launched deviceQuery);
    3. \n
  3. Python, Git, MinnGW, Cygwin installation;
    4. \n
  4. Modified the msvc9compiler.py file and added the line: ld_args.append('/MANIFEST')
    5. \n
  5. PyCUDA installation doesn't work!
    6. \n
\n\n\n

BOOST_INC_DIR = []
BOOST_LIB_DIR = []
BOOST_COMPILER = 'gcc43'
USE_SHIPPED_BOOST = True
BOOST_PYTHON_LIBNAME = ['boost_python-py27']
BOOST_THREAD_LIBNAME = ['boost_thread']
CUDA_TRACE = False
CUDA_ROOT = 'C:\\\\Program Files\\\\NVIDIA GPU Computing Toolkit\\\\CUDA\\\\v6.5'
CUDA_ENABLE_GL = False
CUDA_ENABLE_CURAND = True
CUDADRV_LIB_DIR = ['${CUDA_ROOT}/lib', '${CUDA_ROOT}/lib/x64']
CUDADRV_LIBNAME = ['cuda']
CUDART_LIB_DIR = ['${CUDA_ROOT}/lib', '${CUDA_ROOT}/lib/x64']
CUDART_LIBNAME = ['cudart']
CURAND_LIB_DIR = ['${CUDA_ROOT}/lib', '${CUDA_ROOT}/lib/x64']
CURAND_LIBNAME = ['curand']
CXXFLAGS = ['/EHsc']
LDFLAGS = ['/FORCE']

\n\n

(my cuda path is C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5\\lib\\x64)

\n

6. then executed the following commands: 

\n

set VS90COMNTOOLS=%VS100COMNTOOLS%
python setup.py build
python setup.py install

\n

I don't receive any error.

\n

7. Created this python file:

\n

# from: http://documen.tician.de/pycuda/tutorial.html
import pycuda.gpuarray as gpuarray
import pycuda.driver as cuda
import pycuda.autoinit
import numpy
a_gpu = gpuarray.to_gpu(numpy.random.randn(4,4).astype(numpy.float32))
a_doubled = (2*a_gpu).get()
print a_doubled
print a_gpu

\n\n

and here there's something wrong (please see the screenshot attached).

\n

Would you give me a hand in understanding how to fix it?

\n

Thank you so much

\n

Cristian

" }, { "id": 77597, "postDate": "2015-05-08T10:28:08.640Z", "rawMarkdown": "", "isDeleted": true }, { "id": 77591, "postDate": "2015-05-08T09:54:47.093Z", "rawMarkdown": "", "isDeleted": true }, { "id": 77590, "postDate": "2015-05-08T09:54:34.627Z", "rawMarkdown": "", "isDeleted": true }, { "id": 77586, "postDate": "2015-05-08T09:15:33.437Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 65473, "author_name": "zero zero", "author_url": "", "post_date": "2015-03-05T04:44:30.150000", "content": "

hi there, many of these cool toolkits have been developed on linux systems.  I recommend installing something like ubuntu and then install/build with only a couple of commands.

\n

I can understand one being apprehensive about committing to a dual boot...which does carry risks.  Grab a 3 year old desktop from the recycling and put a used gpu in it (gtx 580 for 100 bucks!) and that'll get you going. Using linux these days is a lot easier than it used to be.

\n

Once you do this, you will never go back!  Pinky-swear!

", "votes": 4, "replies": [] }, { "id": 77595, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-05-08T10:17:25.193000", "content": "

wait  a minute... The important error message here seems to be: 

\n

nvcc fatal : redefinition of argument 'machine'

\n

before reinstalling check it on google! 

\n

----------------------------------------------------------------------------------------------------------------------------------

\n\n

- pycuda is not necessary for theano working

\n

- I wrote microsoft visual studio 12 + cuda 6.5

\n

- to install theano:

\n", "votes": 1, "replies": [] }, { "id": 77574, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-05-08T06:21:48.863000", "content": "

Hello Mohammad,

\n

You get the error not because of that file missing, which indeed is in your file system.

\n

Try not to use that line (flags='-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include') in your .theanorc.txt) and check if it works.

\n

It's a real pain.... :-)

", "votes": 1, "replies": [] }, { "id": 65768, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-09T10:34:13.543000", "content": "

It was a pain, but at the end I've got Theano installed on a Windows 64-bit laptop.

\n

I had to uninstall and re-install Visual Studio, because of the damn crtdefs.h missing.

\n

Ciao!

\n

Cristian

", "votes": 1, "replies": [] }, { "id": 65533, "author_name": "Gregory Godin", "author_url": "", "post_date": "2015-03-05T19:54:45.027000", "content": "

Well, i was able to do gpu on windows (nolearn, lasagne and theano), but that was not easy.

\n

I used msys + mingw32 + win-builds (http://win-builds.org/1.5.0/msys-cygwin.html#_msys_installation)

\n

then installed cuda 6.5 & python2.7

\n

For theano, i used the prebuilds packages from here : http://www.lfd.uci.edu/~gohlke/pythonlibs/#pil

\n

I had to install pthread port : https://sourceware.org/pthreads-win32/

\n

I had to change some python files (didn't write it up and don't remember which files, but i think it was to configure properly pthread in theano ).

\n

And finally, there is a glitch with pylearn2 + cuda convnet.

\n

I had to add cuda_ndarray.pyd and libcuda_ndarray.pyd (got the two files from a tmp folder when python is compiling the cuda_convnet wrapper) to a folder in my PATH env variable.

\n

Hope this will help ! :)

\n\n

Edit : btw, here is my theanorc : 

\n

[global]
device = gpu1
floatX = float32
nocleanup = True

\n

[nvcc]
flags=-LC:\\Python27\\libs
compiler_bindir=C:\\Program Files (x86)\\Microsoft Visual Studio 10.0\\VC\\bin
fastmath = True

", "votes": 1, "replies": [] }, { "id": 65507, "author_name": "inversion", "author_url": "", "post_date": "2015-03-05T13:29:34.737000", "content": "

[quote=zero zero;65473]

\n

I can understand one being apprehensive about committing to a dual boot...which does carry risks. 

\n

[/quote]

\n

I finally gave up a few weeks ago trying to do GPU on windows. My solution was to install Ubuntu on a completely different and brand-new drive of my desktop. To switch environments, I just flip the boot priority when restarting.

\n

It's worked great.

", "votes": 2, "replies": [] }, { "id": 136538, "author_name": "KharlanovSergey", "author_url": "", "post_date": "2016-09-22T18:41:15.547000", "content": "

http://web.archive.org/web/20150601112123/http://machinelearning.berlin/

", "votes": 0, "replies": [] }, { "id": 135063, "author_name": "djab", "author_url": "", "post_date": "2016-09-10T17:00:21.877000", "content": "

[quote=Christian Schröder de Witt;80419]

\n

Hi, try this tutorial here, it's about installing Theano on Windows 8.1 x64:

\n

http://machinelearning.berlin/?p=383

\n

I wrote it because I couldn't find anything that worked. If you stick exactly to the tutorial, it should work (you have few choices though).

\n

All the best Christian

\n

[/quote]

\n

Anyone knows if the tutorial is still availible somewhere else?

", "votes": 0, "replies": [] }, { "id": 80419, "author_name": "Christian Schröder de Witt", "author_url": "", "post_date": "2015-05-29T21:23:00.453000", "content": "

Hi, try this tutorial here, it's about installing Theano on Windows 8.1 x64:

\n

http://machinelearning.berlin/?p=383

\n

I wrote it because I couldn't find anything that worked. If you stick exactly to the tutorial, it should work (you have few choices though).

\n

All the best Christian

", "votes": 0, "replies": [] }, { "id": 77589, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-05-08T09:42:21.273000", "content": "

But first check one more time your theanorc file.

\n

Does it have this line?

\n

[cuda]

\n

root=your_path_to_cuda (v7.0)

\n

root=C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5 <--

", "votes": 0, "replies": [] }, { "id": 77588, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-05-08T09:37:14.700000", "content": "

Not being an expert, I'd suggest to unistall cuda 7.0 and install cuda 6.5, which is a version which has been demonstrated to work good on my windows laptop :-) together with Microsoft Visual Studio 12.0.

\n

Happy fighting :-)

\n

Let me know about it!

", "votes": 0, "replies": [] }, { "id": 77506, "author_name": "mohammadahmed", "author_url": "", "post_date": "2015-05-07T18:42:36.893000", "content": "

[quote=Buzz Lightyear;65768]

\n

It was a pain, but at the end I've got Theano installed on a Windows 64-bit laptop.

\n

I had to uninstall and re-install Visual Studio, because of the damn crtdefs.h missing.

\n

Ciao!

\n

Cristian

\n

[/quote]

Hello Buzz, i read this post for many times until i got to your last problem :

ValueError: Theano nvcc.flags support only parameter/value pairs without space between them. e.g.: '--machine 64' is not supported, but '--machine=64' is supported. Please add the '=' symbol. nvcc.flags value is ''-ID:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include''

So i check for the missing file crtdefs.h, but it was on my system actually! And i have still the last problem what you suggest

", "votes": 0, "replies": [] }, { "id": 73189, "author_name": "lucas", "author_url": "", "post_date": "2015-04-22T15:51:47.517000", "content": "

Ciao Cristian,

\n

I did that already and placed the file in the proper dir, my problem is that I don't know how to replace that specific row with a row customized for Anaconda. Do you mean that in Anaconda that call is not needed so I can basically drop it or should I replace it? (therefore my question, I wouldn't know how)
As for now I'll try proceeding that way but I feel clumsier as I move on, lots of uncertainty, it's not my specialty I must admit.

\n

Thanks for you time,

\n

ciao,

\n

Luca

", "votes": 0, "replies": [] }, { "id": 73173, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-04-22T14:19:20.880000", "content": "

Ciao Luca,

\n

That call is aimed at add winpython stuff. But you don't have to configure that file (it's located here C:\\SciSoft\\WinPython-64bit-2.7.9.3\\scripts\\env.bat on my laptop). You should configure an other env.bat (which is located here "C:\\Scisoft\\env.bat" on my laptop) as I've mentioned above.

\n

Fammi sapere

\n

Ciao

\n

Cristian

", "votes": 0, "replies": [] }, { "id": 73144, "author_name": "lucas", "author_url": "", "post_date": "2015-04-22T11:02:36.747000", "content": "

Hi Buzz Lightyear,

\n

I’m trying to install Theano on my laptop as well, struggling quite a bit, maybe you can help. I’m using Anaconda and I’m following step by step the guide provided here: http://theano.readthedocs.org/en/latest/install_windows.html. The same process you have been following. I’m stuck with the environment configuration script 'env.bat' needed to configure the system path. The tutorial example refers to a WinPython distribution but as I’m installing in Anaconda and I don’t know how to configure that following row:

\n

CALL %SCISOFT%\\WinPython-64bit-2.7.9.4\\scripts\\env.bat

\n

 What location and file should I call in Anaconda (or since you are using WinPython) what is that Call aimed at in WinPython? so that I can try finding the same thingh in anaconda.

\n

Thanks,

\n

Saluti,

\n

Luca

", "votes": 0, "replies": [] }, { "id": 65575, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-06T08:06:10.910000", "content": "

Hi Gregory,

\n

If I had that line:

\n

flags=-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include

\n

or

\n

flags='-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include'

\n

to theanorc I got an other error message:

\n

ValueError: Theano nvcc.flags support only parameter/value pairs without space between them. e.g.: '--machine 64' is not supported, but '--machine=64' is supported. Please add the '=' symbol. nvcc.flags value is '-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include'

\n

Any other help/hint?

\n

Thank you

\n

Cristian

", "votes": 0, "replies": [] }, { "id": 65547, "author_name": "Gregory Godin", "author_url": "", "post_date": "2015-03-05T21:23:14.790000", "content": "

In your theanorc, i gess something like this should do the trick :

\n

[nvcc]

\n

flags=-IC:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include

\n\n

Or you could add the visual include folder to your PATH (ugly)

", "votes": 0, "replies": [] }, { "id": 65545, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-05T21:10:07.777000", "content": "

Hi Gregory, thank you for your help!

\n

Could you be even more clear please? :-)

\n

How do I add that path as include path?

", "votes": 0, "replies": [] }, { "id": 65543, "author_name": "Gregory Godin", "author_url": "", "post_date": "2015-03-05T20:54:04.723000", "content": "

Yes, it can't find "crtdefs.h". Either add the following path as include path : "C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\include" (-I option), either use vcvarsall to properly setup your command prompt.

", "votes": 0, "replies": [] }, { "id": 65539, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-05T20:27:22.230000", "content": "

It's just a pain.... Now theano works but doesn't use the gpu. Arghhhhh!!!

\n

This is my laptop config is x64, nvcc/cuda 6.5, Python 2.7.9, WinPython-64bit-2.7.9.3, Windows 8.1, VS2013 and two graphic units (Intel HD Graphics 4600 and NVIDIA GeForce GT 750M).

\n

where gcc:
C:\\SciSoft\\TDM-GCC-64\\bin\\gcc.exe
C:\\Rtools\\gcc-4.6.3\\bin\\gcc.exe
C:\\MinGW\\bin\\gcc.exe

\n

where gendef:
C:\\SciSoft\\TDM-GCC-64\\x86_64-w64-mingw32\\bin\\gendef.exe

\n

where cl:
C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0\\VC\\bin\\amd64\\cl.exe
C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64\\cl.exe

\n

where nvcc:
C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5\\bin\\nvcc.exe

\n

I did a manual installation. Got the latest bleeding edge version GitHub and downloaded the latest zip. The .theanorc.txt file is located in C:\\SciSoft\\WinPython-64bit-2.7.9.3\\settings:

\n

[global]
device=gpu
floatX=float32

\n

[cuda]
root=C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5

\n

[nvcc]
fastmath=True
compiler_bindir=C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64
cxx=C:\\SciSoft\\TDM-GCC-64\\bin

\n

[blas]
ldflags=

\n

And here is what I get from a test....

\n

c:\\program files\\nvidia gpu computing toolkit\\cuda\\v6.5\\include\\host_config.h(149) : fatal error C1083: Impossibile aprire il file inclusione 'crtdefs.h': No such file or directory ERROR (theano.sandbox.cuda): Failed to compile cuda_ndarray.cu: ('nvcc return status', 2, 'for cmd', 'nvcc -shared -O3 --compiler-bindir C:\\\\Program Files (x86)\\\\Microsoft Visual Studio 12.0\\\\VC\\\\bin\\\\amd64 -Xlinker /DEBUG -D HAVE_ROUND -m64 -Xcompiler -DCUDA_NDARRAY_CUH=7772254ca6db84d79e28ecddbe019d53,-D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD -Ic:\\\\scisoft\\\\theano-master\\\\theano\\\\sandbox\\\\cuda -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\lib\\\\site-packages\\\\numpy\\\\core\\\\include -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\include -o C:\\\\Users\\\\Cris\\\\AppData\\\\Local\\\\Theano\\\\compiledir_Windows-8-6.2.9200-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.9-64\\\\cuda_ndarray\\\\cuda_ndarray.pyd mod.cu -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\libs -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64 -lpython27 -lcublas -lcudart') ERROR:theano.sandbox.cuda:Failed to compile cuda_ndarray.cu: ('nvcc return status', 2, 'for cmd', 'nvcc -shared -O3 --compiler-bindir C:\\\\Program Files (x86)\\\\Microsoft Visual Studio 12.0\\\\VC\\\\bin\\\\amd64 -Xlinker /DEBUG -D HAVE_ROUND -m64 -Xcompiler -DCUDA_NDARRAY_CUH=7772254ca6db84d79e28ecddbe019d53,-D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD -Ic:\\\\scisoft\\\\theano-master\\\\theano\\\\sandbox\\\\cuda -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\lib\\\\site-packages\\\\numpy\\\\core\\\\include -IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\include -o C:\\\\Users\\\\Cris\\\\AppData\\\\Local\\\\Theano\\\\compiledir_Windows-8-6.2.9200-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.9-64\\\\cuda_ndarray\\\\cuda_ndarray.pyd mod.cu -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\libs -LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64 -lpython27 -lcublas -lcudart') WARNING (theano.sandbox.cuda): CUDA is installed, but device gpu is not available (error: cuda unavilable) WARNING:theano.sandbox.cuda:CUDA is installed, but device gpu is not available (error: cuda unavilable)

\n

mod.cu

\n

['nvcc', '-shared', '-O3', '-use_fast_math', '--compiler-bindir', 'C:\\\\Program Files (x86)\\\\Microsoft Visual Studio 12.0\\\\VC\\\\bin\\\\amd64', '-Xlinker', '/DEBUG', '-D HAVE_ROUND', '-m64', '-Xcompiler', '-DCUDA_NDARRAY_CUH=7772254ca6db84d79e28ecddbe019d53,-D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD', '-Ic:\\\\scisoft\\\\theano-master\\\\theano\\\\sandbox\\\\cuda', '-IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\lib\\\\site-packages\\\\numpy\\\\core\\\\include', '-IC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\include', '-o', 'C:\\\\Users\\\\Cris\\\\AppData\\\\Local\\\\Theano\\\\compiledir_Windows-8-6.2.9200-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.9-64\\\\cuda_ndarray\\\\cuda_ndarray.pyd', 'mod.cu', '-LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64\\\\libs', '-LC:\\\\SciSoft\\\\WinPython-64bit-2.7.9.3\\\\python-2.7.9.amd64', '-lpython27', '-lcublas', '-lcudart']
NP time: 0.116000[s], theano time: 0.114000[s] (times should be close when run on CPU!)
Result difference: 0.000000

\n

Does anybody know why it doesn't use the gpu? Cuda is installed and pass the deviceQuery test!

\n

I've been spending days on it....

\n

Thank you for your help

\n

Cristian

", "votes": 0, "replies": [] }, { "id": 65538, "author_name": "Richard Giles", "author_url": "", "post_date": "2015-03-05T20:15:59.717000", "content": "

I wasted countless hours/days trying to get this working on Windows 8.1. I finally gave up on this. It is just not very well supported. Best to stick with Linux distributions with and without GPU support.

", "votes": 0, "replies": [] }, { "id": 65496, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-05T12:06:17.977000", "content": "

Hi,

\n

Here Theano is still not working....
This is my laptop config is x64, nvcc/cuda 6.5, Python 2.7.9, WinPython-64bit-2.7.9.3, Windows 8.1, VS2013 and two graphic units (Intel HD Graphics 4600 and NVIDIA GeForce GT 750M).

\n

This is my env.bat in c:\\SciSoft

\n

REM configuration of paths
set VSFORPYTHON="C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0"
set SCISOFT=%~dp0REM add tdm gcc stuff
set PATH=%SCISOFT%\\TDM-GCC-64\\bin;%SCISOFT%\\TDM-GCC-64\\x86_64-
w64-mingw32\\bin;%PATH%REM add winpython stuff
CALL %SCISOFT%\\WinPython-64bit-2.7.6.4\\scripts\\env.batREM configure path for msvc compilers
REM for a 32 bit installation change this line to
REM CALL %VSFORPYTHON%\\vcvarsall.bat
CALL %VSFORPYTHON%\\vcvarsall.bat amd64REM return a shell
cmd.exe /k
where gcc:
C:\\SciSoft\\TDM-GCC-64\\bin\\gcc.exe
C:\\Rtools\\gcc-4.6.3\\bin\\gcc.exe
C:\\MinGW\\bin\\gcc.exe

\n

where gendef:
C:\\SciSoft\\TDM-GCC-64\\x86_64-w64-mingw32\\bin\\gendef.exe

\n

where cl:
C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0\\VC\\bin\\amd64\\cl.exe
C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64\\cl.exe

\n

where nvcc:
C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v6.5\\bin\\nvcc.exe

\n

I uninstalled and re-installed Theano from github following the instructions.
The .theanorc.txt file is located in C:\\SciSoft\\WinPython-64bit-2.7.9.3\\settings:

\n


[global]
device = gpu
floatX = float32

\n

[nvcc]

\n

compiler_bindir=C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin\\amd64
cxx = C:\\SciSoft\\TDM-GCC-64\\bin
flags = --use-local-env --cl-version=2008

\n

But still not working.
Theano seems installed.... the main error message is "global name 'CVM' is not defined"

\n

Please check the notebook enclosed...
Thank you very much for your help

\n

Regards
Cristian

", "votes": 0, "replies": [] }, { "id": 65421, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-04T18:01:44.993000", "content": "

I tried an other way, following http://deeplearning.net/software/theano/install_windows.html#install-windows.

\n

1- Visual Studio 2013 and Cuda 6.5 installed.

\n

2- Install Microsoft Visual C++ Compiler for Python 2.7

\n\n

3- Install TDM GCC in C:\\SciSoft\\TDM-GCC-64

\n

4- Install WinPython in C:\\SciSoft\\WinPython-64bit-2.7.9.3

\n

5- Configuring the environment. Saved the following shell script into c:\\scisoft\\env.bat to configure the system path:

\n

REM configuration of paths
set VSFORPYTHON="C:\\Program Files (x86)\\Common Files\\Microsoft\\Visual C++ for Python\\9.0"
set SCISOFT=%~dp0

REM add tdm gcc stuff
set PATH=%SCISOFT%\\TDM-GCC-64\\bin;%SCISOFT%\\TDM-GCC-64\\x86_64-w64-mingw32\\bin;%PATH%

REM add winpython stuff
CALL %SCISOFT%\\WinPython-64bit-2.7.9.3\\scripts\\env.bat

REM configure path for msvc compilers
REM for a 32 bit installation change this line to
REM CALL %VSFORPYTHON%\\vcvarsall.bat
CALL %VSFORPYTHON%\\vcvarsall.bat amd64

REM return a shell
cmd.exe /k

\n

6- create a link library for GCC. I opened the WinPython Command Prompt from C:\\SciSoft\\WinPython-64bit-2.7.9.3 and executed:

\n

gendef WinPython-64bit-2.7.9.3\\python-2.7.9.amd64\\python27.dll
dlltool --dllname python27.dll --def python27.def --output-lib WinPython-64bit-2.7.9.3\\python-2.7.9.amd64\\libs\\libpython27.a

\n


7- install Theano from github in C:\\SciSoft\\Theano

\n

8- from C:\\SciSoft\\Theano

\n

python setup.py develop

\n

9- configure Theano for GPU use (.theanorc.txt file) placed in C:\\scisoft\\WinPython-64bit-2.7.9.3\\settings

\n

[global]
device = gpu
floatX = float32

[nvcc]
flags = --use-local-env --cl-version=2008

\n

10- run a test http://deeplearning.net/software/theano/tutorial/using_gpu.html#using-gpu

\n

I got an Error Message... please check the attached file....

\n\n

Thank you so much for your help!

\n

I won't give up :-)

", "votes": 0, "replies": [] }, { "id": 65410, "author_name": "Buzz Lightyear", "author_url": "", "post_date": "2015-03-04T14:20:20.243000", "content": "

@ sedielem: many thanks, but still in big troubles... As I wrote before cuda 6.5 is installed.

\n

Now I'm trying to install theano. I open Git Bash and I clone from github.

\n

git clone git://github.com/Theano/Theano.git

\n

Then from the folder where Theano files are:

\n

python setup.py install

\n

I got this long message error (please check the attached file).

\n

That's a nightmare...

\n

Thank you for your help!

", "votes": 0, "replies": [] }, { "id": 65403, "author_name": "sedielem", "author_url": "", "post_date": "2015-03-04T12:03:45.117000", "content": "

you do not need PyCUDA unless you plan to use the FFT-based convolution implementation. As far as I'm aware there are no other components in Theano right now that depend on PyCUDA.

\n

If you want to install it anyway, I'm afraid I can't help you with that - but it shouldn't be necessary :)

", "votes": 0, "replies": [] }, { "id": 77597, "author_name": "", "author_url": "", "post_date": "2015-05-08T10:28:08.640000", "content": "", "votes": 0, "replies": [] }, { "id": 77591, "author_name": "", "author_url": "", "post_date": "2015-05-08T09:54:47.093000", "content": "", "votes": 0, "replies": [] }, { "id": 77590, "author_name": "", "author_url": "", "post_date": "2015-05-08T09:54:34.627000", "content": "", "votes": 0, "replies": [] }, { "id": 77586, "author_name": "", "author_url": "", "post_date": "2015-05-08T09:15:33.437000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "65473": "", "77595": "", "77574": "", "65768": "", "65533": "", "65507": "", "136538": "http://web.archive.org/web/20150601112123/http://machinelearning.berlin/", "135063": "", "80419": "", "77589": "", "77588": "", "77506": "", "73189": "", "73173": "", "73144": "", "65575": "", "65547": "", "65545": "", "65543": "", "65539": "", "65538": "", "65496": "", "65421": "", "65410": "", "65403": "", "65396": "", "77597": "", "77591": "", "77590": "", "77586": "" } }