I am new to tensorflow and I am working on shared linux (Ubuntu 16.04), it means I don't have root access. Cuda 8.0 and Cudnn 8 are already installed by admin as root. I have installed python 3.5 using anaconda and then installed tensorflow using pip. I have added the cuda-8.0/bin and cuda-8.0/lib64 to PATH and LD_PATH_LIBRARY using following exports.
export PATH="$PATH:/usr/local/cuda-8.0/bin"
export LD_LIBRARY_PATH="/usr/local/cuda-8.0/lib64"
But when I try to run the program it gives the following error.
ImportError: libcublas.so.8.0: cannot open shared object file: No such file or directory
However these files exist in LD_LIBRARY_PATH, and nvcc -V is also working.
Is it even possible to refer to the system installed Cuda and CuDnn ? If yes, can you help to clear the above error. Thanks in advance.
A likely explanation is that your path is not set up correctly. Try echo $LD_LIBRARY_PATH and let us know what you get.
Another explanation is that it is not in that directory. Yes, libcublas.so should normally be in /usr/local/cuda-8.0/lib64 but double check if it is there or another directory by using find.
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.