ImportError: libcudart.so.7.0: cannot open shared object file: No such file or directory

[monajalal]

I get the following error when I run the sample:

mona@pascal:~/computer_vision/py-faster-rcnn$ ./tools/demo.py
Traceback (most recent call last):
  File "./tools/demo.py", line 18, in <module>
    from fast_rcnn.test import im_detect
  File "/home/mona/computer_vision/py-faster-rcnn/tools/../lib/fast_rcnn/test.py", line 17, in <module>
    from fast_rcnn.nms_wrapper import nms
  File "/home/mona/computer_vision/py-faster-rcnn/tools/../lib/fast_rcnn/nms_wrapper.py", line 9, in <module>
    from nms.gpu_nms import gpu_nms
ImportError: libcudart.so.7.0: cannot open shared object file: No such file or directory

I have the following system settings:

CuDNN V4
$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2015 NVIDIA Corporation
Built on Tue_Aug_11_14:27:32_CDT_2015
Cuda compilation tools, release 7.5, V7.5.17
$ uname -a
Linux pascal 3.13.0-62-generic #102-Ubuntu SMP Tue Aug 11 14:29:36 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux
$ lspci | grep -i nvidia
03:00.0 3D controller: NVIDIA Corporation GK110BGL [Tesla K40c] (rev a1)
83:00.0 3D controller: NVIDIA Corporation GK110BGL [Tesla K40c] (rev a1)
$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 14.04.5 LTS
Release:    14.04
Codename:   trusty
$ gcc --version
gcc (Ubuntu 4.8.4-2ubuntu1~14.04.3) 4.8.4
Copyright (C) 2013 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

What is the issue and how can it be solved?

[monajalal]

In my case I changed the paths to use the already installed Cuda7.0. Seems Cuda7.5 is not compatible with Ubuntu 14.04?

[D3Nd3R]

I think you should add the next lines to the .bashrc file. But I did it on Ubuntu 16.04 and cuda 8.0.
export PATH=/usr/local/cuda-7.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-7.0/lib64${LD_LIBRARY_PATH:+: ${LD_LIBRARY_PATH}}

[ravikantb]

I also faced the similar issue with CUDA 8.0 not being detected and the training code instead looking for older CUDA 7.5. I updated my PATH and LD_LIBRARY_PATH flags but it didn't resolve. Something more was needed. The hint lied in the error log itself.

The line: from nms.gpu_nms import gpu_nms is from 'nms_wrapper.py' which is in 'lib/fast_rcnn' directory of faster-RCNN module. The same 'lib' directory also contains 'nms/gpu_nms.so' which in my case was built against older CUDA 7.5. And whenever I did 'make' in 'lib' directory after updating my CUDA to 8.0 (as per the instruction in README), nothing was being done as older copy was already present. I removed all the '*.so' files in 'lib' directory and did a 'make' again. This time new *.so files were created using CUDA 8.0 (you can verify that from the build logs also).

Now my training is running smoothly (so far, fingers crossed!).

[eswears]

This fixed the problem for me:
sudo ldconfig /usr/local/cuda/lib64