Failed to load native TensorFlow runtime

General Discussion
#1

Hi All,
I am new to TensorFlow - this is the stack trace of my error

 File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/home/megh/anaconda3/envs/bne/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/megh/anaconda3/envs/bne/lib/python3.6/imp.py", line 243, in load_module
    return load_dynamic(name, filename, file)
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/imp.py", line 343, in load_dynamic
    return _load(spec)
ImportError: libcublas.so.9.0: 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 "bne.py", line 21, in <module>
    from encoder import *
  File "/home/megh/projects/entity-norm/BNE/encoder.py", line 2, in <module>
    import tensorflow as tf
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/__init__.py", line 24, in <module>
    from tensorflow.python import pywrap_tensorflow  # pylint: disable=unused-import
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/python/__init__.py", line 49, in <module>
    from tensorflow.python import pywrap_tensorflow
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow.py", line 74, in <module>
    raise ImportError(msg)
ImportError: Traceback (most recent call last):
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/home/megh/anaconda3/envs/bne/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/megh/anaconda3/envs/bne/lib/python3.6/imp.py", line 243, in load_module
    return load_dynamic(name, filename, file)
  File "/home/megh/anaconda3/envs/bne/lib/python3.6/imp.py", line 343, in load_dynamic
    return _load(spec)
ImportError: libcublas.so.9.0: cannot open shared object file: No such file or directory


Failed to load the native TensorFlow runtime.

See https://www.tensorflow.org/install/errors

for some common reasons and solutions.  Include the entire stack trace
above this error message when asking for help.

These are my system specifications

  Operating System: Ubuntu 20.04.1 LTS
            Kernel: Linux 5.8.0-36-generic
      Architecture: x86-64

My tensorflow versions are -

tensorboard               1.12.2                   pypi_0    pypi
tensorflow                1.12.0                   pypi_0    pypi
tensorflow-gpu            1.12.0                   pypi_0    pypi

I am running the codes on an NVIDIA GeForce RTX 2080 and CUDA 11.0

Am I facing this issue due to the mismatch between my TensorFlow version and CUDA version?

Any help will greatly be appreciated.

Thank you in advance,
Megh

#3

Hi Megh,
Please take a look at tested build configurations for pre built TF binaries.
Provided your system configuration supports AVX instructions sets, I suspect that you are having TF and Cuda incompatibility issue.
TF 2.4 is compatible with Cuda 11.0 as installed in your system.
Currently you see error because TF is looking for cuda 9.0 which is compatible with TF 1.12
I would encourage you to switch to TF 2.X versions since they are actively supported by TF team.
Hope this helps. Thanks!

#5

Hi @Yasir_Modak - thank you so much for your response and apologies for my delayed reply.
Assuming I want to use tensorflow 1.12.0 itself, I can install cudatoolkit 9 in my virtual environment, right?

#7

That’s correct. You should install cuda 9.0