Possible source of <cuda_runtime_api.h> error on Windows

[N2ITN]

I am getting the following error on compilation:

In file included from src/nnet/conv_bc01_matmul.cpp:1:0:
./include/cudarray/common.hpp:8:30: fatal error: cuda_runtime_api.h: No such file or directory
 #include <cuda_runtime_api.h>

I suspect the reason is that by default, Windows installs CUDA a path that includes spaces: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v7.5/

I know CUDA is installed properly on my system because it works fine using cuDNN with Keras.

If this is the case, is there a workaround, like providing the make file with a link?

Changing the file path might work, but it seems like a stretch.

[JackYoustra]

Alternatively, perhaps it could work if a symlink was created

[lborg019]

same error, no idea how to fix. According to this stackoverflow question, it MIGHT be solvable using pyCUDA and Conda.

http://stackoverflow.com/questions/38293639/errors-installing-cuda-back-end-or-cudarray-in-python-in-windows-10

[KeyWgh]

Have you issued that problem yet? It really cost me a lot of time to find a solution.

[wickedcode01]

the same problem

[wickedcode01]

we need the provider fix it or give us some solution!

[wickedcode01]

@andersbll