ImportError: libcudnn.so.6: cannot open shared obj

Question

I get following error while importing Tensorflow.

    >>> import tensorflow
Traceback (most recent call last):
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow.py", line 41, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 24, in swig_import_helper
    _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
  File "/home/jarvis/anaconda3/lib/python3.6/imp.py", line 242, in load_module
    return load_dynamic(name, filename, file)
  File "/home/jarvis/anaconda3/lib/python3.6/imp.py", line 342, in load_dynamic
    return _load(spec)
ImportError: libcudnn.so.6: cannot open shared object file: No such file or directory

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/__init__.py", line 24, in <module>
    from tensorflow.python import *
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/__init__.py", line 49, in <module>
    from tensorflow.python import pywrap_tensorflow
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow.py", line 52, in <module>
    raise ImportError(msg)
ImportError: Traceback (most recent call last):
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow.py", line 41, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/home/jarvis/anaconda3/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 24, in swig_import_helper
    _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
  File "/home/jarvis/anaconda3/lib/python3.6/imp.py", line 242, in load_module
    return load_dynamic(name, filename, file)
  File "/home/jarvis/anaconda3/lib/python3.6/imp.py", line 342, in load_dynamic
    return _load(spec)
ImportError: libcudnn.so.6: cannot open shared object file: No such file or directory


Failed to load the native TensorFlow runtime.

I added the path variables too

$ export PATH=/usr/local/cuda-8.0/bin${PATH:+:${PATH}}
$ export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

Then the system detects nvcc. However as soon as close the terminal windows, and type nvcc -V in a new terminal, it again shows not installed

Answer 1

actually, I met this problem in installing tensorflow-gpu according the official installing document. I solved it by changing the version of tensorflow-gpu.

pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.3.0-cp36-cp36m-linux_x86_64.whl

I changed the number 1.3.0 to 1.2.0 ,it works.the new install commend is below:

pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.2.0-cp36-cp36m-linux_x86_64.whl

Answer 2

You need to install cuDNN v6.0 library.

Please refer to the following link.

After building TensorFlow from source, seeing libcudart.so and libcudnn errors

Answer 3

One thing that worked for me was that rather than using 'export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64" ' I used 'export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:~/cuda/lib64" ' and it worked. I have to do it everytime the computer starts. Dont know what is cuda doing in my home folder and don't know why I need to do it everytime.