Reputation: 65
I am trying to use Theano for Deep Learning in Windows 7. I have installed Visual Studio Community Edition 2013 . And CUDA Toolkit 7.5 . I am using Anaconda environment for Python.
But when I import Theano I encounter the following command line message.
ERROR:theano.sandbox.cuda:Failed to compile cuda_ndarray.cu: ('nvcc return statu s', 2, 'for cmd', 'nvcc -shared -O3-LC:\\Users\\hp\\Anaconda2\\libs -use_fast_m ath --compiler-bindir C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\ \bin\\amd64-Xlinker /DEBUG -D HAVE_ROUND -m64 -Xcompiler -DCUDA_NDARRAY_CUH=187 15462c72ed6afcd7ca5d52813ce90,-DNPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION,/Zi,/MD -IC:\\Users\\hp\\Anaconda2\\lib\\site-packages\\theano\\sandbox\\cuda -IC:\\Us ers\\hp\\Anaconda2\\lib\\site-packages\\numpy\\core\\include -IC:\\Users\\hp\\An aconda2\\include -IC:\\Users\\hp\\Anaconda2\\lib\\site-packages\\theano\\gof -o C:\\Users\\hp\\AppData\\Local\\Theano\\compiledir_Windows-7-6.1.7601-SP1-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.12-64\\cuda_ndarray\\cuda_ndarray.pyd mod.cu -LC:\\Users\\hp\\Anaconda2\\libs-LC:\\Users\\hp\\Anaconda2 -lcublas -lpython27 -lcudart')
My Config file(.theanorc) looks like this::
[global]
floatX = float32
openmp =false
device = gpu
mode=FAST_RUN
[cuda]
root = C:\Program Files\NVIDIA Corporation\Installer2\CUDAToolkit_7.5
[nvcc]
flags = -LC:\Users\hp\Anaconda2\libs
compiler_bindir = C:\Program Files (x86)\Microsoft Visual Studio12.0\VC\bin\amd64
fastmath = True
Please help
Upvotes: 0
Views: 97
Reputation: 26
I am not sure if this is the error, but you should change this line:
flags = -LC:\Users\hp\Anaconda2\libs
It is looking in -LC:\Users\hp\Anaconda2\libs
instead of C:\Users\hp\Anaconda2\libs
.
Upvotes: 1
Reputation: 60321
You should use quotes in directories; also, I had similar errors, until I removed the compliler_bindir
line. I suggest you start with this minimal .theanorc
file:
[global]
floatX = float32
device = gpu
[nvcc]
flags=-L"C:\Users\hp\Anaconda2\libs"
provided of course that the flags
path is the correct one...
Upvotes: 0