cudarray icon indicating copy to clipboard operation
cudarray copied to clipboard

Published 20 hours ago • verified publisher icon andersbll

CUDArray: CUDA back-end not available, using NumPy

Open 00fq00 opened this issue 8 years ago • 5 comments

I've installed cudarray, But Still says CUDArray: CUDA back-end not available, using NumPy,I'm sure my rLD_LIBRARY_PATH points to libcudarray.so ,but it doesn't work

PATH=/usr/local/cuda-7.5/bin:/usr/local/bin

Python 2.7.5 (default, Aug  9 2016, 05:27:46) 
[GCC 4.8.5 20150623 (Red Hat 4.8.5-4)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import cudarray
CUDArray: CUDA back-end not available, using NumPy.
>>>

00fq00 avatar Sep 02 '16 06:09 00fq00

Hi, @00fq00 I think this is because incorrect cudarray installation. Please include below to /etc/profile and clone a new cudarray repo and reinstall it.

export INSTALL_PREFIX=/usr/local export CUDA_PREFIX=/usr/local/cuda export CUDNN_ENABLED=1

Thanks,

oscarriddle avatar Sep 11 '16 08:09 oscarriddle

@oscarriddle Hello, I had this same issue and unfortunately changing the environment variables didn't fix it.

mxchinegod avatar Oct 05 '16 18:10 mxchinegod

I have the same issue as lots of people have had. I thought that the problem might be that the CUDArray build assumes that the CUDA libraries are available in /usr/local/cuda/lib. On my setup there is no such folder. I symlinked /usr/local/cuda/lib64 to /usr/local/cuda/lib and rebuilt CUDArrary.

But that did not fix the problem. CUDArray is just proxying through to NymPy.

I'm setting INSTALL_PREFIX, CUDA_PREFIX, CUDNN_ENABLED, and even LD_LIBRARY_PATH. Doesn't work. No indication about why until you run it and it proxies through to NumPy.

endymion avatar Oct 29 '16 17:10 endymion

I am running into the same issue. Has anyone been able to resolve it?

bradasteiner avatar Jul 26 '18 18:07 bradasteiner

Hi, I know this is a fairly old post but I am trying to implement the same work you did by installing CUDArray but unfortunately couldn't. I am using pycharm to run my python projects on a windows based machine and after running in C:\user\anaconda3 - python setup.py install it runs totally fine. Then I install deeppy in C:\user\anaconda3 folder again, runs well. But when it comes to running celeba-aegan.py, I get the message. CUDA-backend not available using numpy.

Now, my problems constitute,

  1. I don't know where to pass those prefix and make info.
  2. Do I have to copy everything both cudarray and deeppy in anaconda3 for the installation.

I have been stuck at this so long and before coming here I did a lot of research to avoid bothering you.

I hope you understand. Thanks

BaqarAbbas55 avatar Nov 24 '20 12:11 BaqarAbbas55