TensorFlow: libcudart.so.7.5: cannot open shared object file: No such file or directory

Question

I am running TensorFlow on Ubuntu 15.10. When I enter pip show tensorflow, I see that TF has been installed properly.

However, when I write import tensorflow as tf, I get the following error message:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/__init__.py", line 23, in <module>
    from tensorflow.python import *
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/python/__init__.py", line 49, in <module>
    from tensorflow import contrib
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/contrib/__init__.py", line 23, in <module>
    from tensorflow.contrib import layers
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/contrib/layers/__init__.py", line 68, in <module>
    from tensorflow.contrib.layers.python.layers import *
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/contrib/layers/python/layers/__init__.py", line 22, in <module>
    from tensorflow.contrib.layers.python.layers.initializers import *
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/contrib/layers/python/layers/initializers.py", line 24, in <module>
    from tensorflow.python.ops import random_ops
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/random_ops.py", line 23, in <module>
    from tensorflow.python.framework import ops
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 39, in <module>
    from tensorflow.python.framework import versions
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/versions.py", line 22, in <module>
    from tensorflow.python import pywrap_tensorflow
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 28, in <module>
    _pywrap_tensorflow = swig_import_helper()
  File "/home/me/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 24, in swig_import_helper
    _mod = imp.load_module('_pywrap_tensorflow', fp, pathname, description)
ImportError: libcudart.so.7.5: cannot open shared object file: No such file or directory

For what it's worth, I have followed the instructions here and set my LD_LIBRARY_PATH and CUDA_HOME environment variables.

Any advice?

EDIT:

I have installed CUDA 7.5 and added these to my .profile file:

export LD_LIBRARY_PATH="/usr/local/cuda-7.5/lib64"
export CUDA_HOME=/usr/local/cuda-7.5

However, I continue to see the same error message.

EDIT:

I see the following output when I run ldd /usr/local/cuda-7.5/lib64/libcudart.so.7.5:

linux-vdso.so.1 =>  (0x00007ffdac7ea000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fc27a281000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fc27a07d000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fc279e5e000)
librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fc279c56000)
/lib64/ld-linux-x86-64.so.2 (0x00005604f5406000)

EDIT:

If it is relevant, I use GeForce GT640.

EDIT:

I followed @tommus' advice and called source ~/.profile before running TensorFlow, and now it works like a charm.

Thanks to everyone who tried to help me in the comments -- this is my first experience with any kind of Linux distribution, and I am really very appreciative of all the patient assistance I have received so far :-) You guys are absolutely awesome!

tensorflow is looking for CUDA 7.5 in this case. Did you install CUDA 7.5 or CUDA 7.0? What path exactly did you set your `LD_LIBRARY_PATH` to? Is there a libcudart.so.7.5 on that path? — Robert Crovella, Mar 22 '16 at 15:57
Hi! I installed CUDA 7.0, and I added this to my .profile file: `export LD_LIBRARY_PATH="/home/me/cuda/targets/ppc64le-linux/lib"` — M.Y. Babt, Mar 22 '16 at 16:01
@RobertCrovella Thanks, I will give it a shot. I will post again if I encounter any more trouble :-) — M.Y. Babt, Mar 22 '16 at 16:06
@RobertCrovella What command should I write to find out the installation folder for CUDA 7.5? — M.Y. Babt, Mar 22 '16 at 16:26
It should be evident from the install process. What process did you use to install it? (for example did you follow the linux installation guide ?) Are you working on a Power8 platform (ppc64le) ? The default install path should be `/usr/local/cuda` but it doesn't appear that you are using default. If you are installing CUDA to your local directory (`/home/me/...`), do you have an appropriate GPU and driver installed? The linux `find` command should be able to find any file, something like `find / -name libcudart.so.7.5`, but this may be a bit cluttered if you are not a root user. — Robert Crovella, Mar 22 '16 at 16:34
@RobertCrovella I have updated my post. I installed CUDA 7.5 but still receive the same error message. — M.Y. Babt, Mar 22 '16 at 16:58
What is the output of the `ldd /usr/local/cuda/lib64/libcudart.so.7.5`? What is the output of the `ls /usr/local/cuda/lib64/`? Of course paths may vary depends on where You have installed CUDA. You may also have to call: `sudo ldconfig` and try to rerun TensorFlow. — Tomasz Dzieniak, Mar 22 '16 at 17:02
However, when I `cd` to the CUDA folder and write `ls`, `libcudart.so.75` is listed. — M.Y. Babt, Mar 22 '16 at 17:05
Are You sure? If the file is there the command `ldd libcudart.so.7.5` must return a value. — Tomasz Dzieniak, Mar 22 '16 at 17:07
Have You sourced `.profile` file after editing it? Call `source ~/.profile` and try to run TensorFlow. — Tomasz Dzieniak, Mar 22 '16 at 17:23
I've added an answer. If it works for You, please mark as valid. Glad I could help. — Tomasz Dzieniak, Mar 22 '16 at 17:29

score 9 · Accepted Answer · answered Mar 22 '16 at 17:29

9

After editing .profile file you either need to log out and log back in or run the following command:

source ~/.profile

The solution is persistent so there is no need to perform it ever again.

answered Mar 22 '16 at 17:29

Tomasz Dzieniak

2,765
3
24
42

Hi, even though your solution worked, I discover that I have to specify `source ~/.profile` every time :-) – M.Y. Babt Mar 22 '16 at 17:53
1

After restarting Your system it should not be necessary. – Tomasz Dzieniak Mar 22 '16 at 17:55

score 5 · Answer 2 · answered Aug 25 '17 at 06:48

5

sudo nano /etc/ld.so.conf.d/nvidia.conf

Add this:

/usr/local/cuda-8.0/lib64

/usr/local/cuda-8.0/lib

run

sudo ldconfig

if it works well ,please tell me.

answered Aug 25 '17 at 06:48

dreamer

59
1
1

This worked well for me too, but please make sure you change the cuda version according to the version you have. for ex. I have 9.0 so the line i had to add was /usr/local/cuda-9.0/lib64. And if you don't know what is your cuda version please look here: https://stackoverflow.com/questions/9727688/how-to-get-the-cuda-version – Neeraj Komuravalli Mar 07 '18 at 21:16
This solved the problem of Pycharm remote interpreter. – Honghe.Wu Feb 21 '19 at 08:38

score 3 · Answer 3 · answered Jan 14 '18 at 22:28

3

If you are using pip but other answers did not work for you, try this.

In my case this exact problem was solved by reinstalling tensorflow using:

pip install tensorflow

Pip must have messed up its dependencies or something when installing other packages after Tensorflow installation (I've run pip install -r requirements.txt afterwards for my own purposes --> that broke Tensorflow).

answered Jan 14 '18 at 22:28

Morphing Coffee

815
1
10
20

1

In my case it was `pip3` – Tim Hallman Feb 26 '18 at 22:29

score 0 · Answer 4 · answered Aug 31 '17 at 19:53

0

You maybe need to update your cuDNN version as Tensorflow 1.3 doesn't work with cuDNN 5.1 or earlier. Download cuDNN 6.0 or upper and install to fix the issue

answered Aug 31 '17 at 19:53

Asad

2,782
2
16
17

score 0 · Answer 5 · edited Jul 16 '18 at 11:02

I had same error. I have to have Cuda 8.0 and CuDNN 5.1 for running some experiments!!

SOLUTION: Run python with Sudo Python

python
Python 2.7.15 |Anaconda, Inc.| (default, May 1 2018, 23:32:55)
[GCC 7.2.0] on linux2
Type "help", "copyright", "credits" or "license" for more information.

import tensorflow
Traceback (most recent call last):
File "", line 1, in 
File "/home/ujjval/anaconda2/lib/python2.7/site-packages/tensorflow/init.py", line 22, in 
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
File "/home/ujjval/anaconda2/lib/python2.7/site-packages/tensorflow/python/init.py", line 49, in 
from tensorflow.python import pywrap_tensorflow
File "/home/ujjval/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 74, in 
raise ImportError(msg)
ImportError: Traceback (most recent call last):
File "/home/ujjval/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in 
from tensorflow.python.pywrap_tensorflow_internal import *
File "/home/ujjval/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in 
_pywrap_tensorflow_internal = swig_import_helper()
File "/home/ujjval/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 24, in swig_import_helper
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
ImportError: libcublas.so.9.0: cannot open shared object file: No such file or directory

Failed to load the native TensorFlow runtime.

TensorFlow: libcudart.so.7.5: cannot open shared object file: No such file or directory

5 Answers5

Linked