0

I'm trying to use Deeplabcut on my work machine, the GPU is a Nvidia NVS 300, driver updated to 342.00. the only Cuda version that does not raise an error is the Cuda Toolkit 8.0 GA2 patch2, so i installed it and then downloaded Cudnn 6.0 (also tried 5.1, should both be okay) and copied the file to the Cuda folders, and set the env system var path to the Cuda bin.

then I moved to Conda, created the env with python 3.6.7 (also tried 3.5 when using Cudnn 5.1) and installed tensorflow-gpu= 1.4.0 (tried every version from 1.1.0 to 1.4.0 changing the Cudnn and python when necessary, all raises same errors).

it all goes well, until I try to import tensorflow to check the version and if the installation was successful. the error is the following:

Traceback (most recent call last):
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 18, in swig_import_helper
    return importlib.import_module(mname)
  File "D:\Francesco\Anaconda\envs\DLC\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 985, in _gcd_import
  File "<frozen importlib._bootstrap>", line 968, in _find_and_load
  File "<frozen importlib._bootstrap>", line 957, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 666, in _load_unlocked
  File "<frozen importlib._bootstrap>", line 577, in module_from_spec
  File "<frozen importlib._bootstrap_external>", line 938, in create_module
  File "<frozen importlib._bootstrap>", line 222, in _call_with_frames_removed
ImportError: DLL load failed: The specified module could not be found.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow.py", line 41, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 21, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 20, in swig_import_helper
    return importlib.import_module('_pywrap_tensorflow_internal')
  File "D:\Francesco\Anaconda\envs\DLC\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
ImportError: No module named '_pywrap_tensorflow_internal'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\__init__.py", line 24, in <module>
    from tensorflow.python import *
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\__init__.py", line 51, in <module>
    from tensorflow.python import pywrap_tensorflow
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow.py", line 52, in <module>
    raise ImportError(msg)
ImportError: Traceback (most recent call last):
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 18, in swig_import_helper
    return importlib.import_module(mname)
  File "D:\Francesco\Anaconda\envs\DLC\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 985, in _gcd_import
  File "<frozen importlib._bootstrap>", line 968, in _find_and_load
  File "<frozen importlib._bootstrap>", line 957, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 666, in _load_unlocked
  File "<frozen importlib._bootstrap>", line 577, in module_from_spec
  File "<frozen importlib._bootstrap_external>", line 938, in create_module
  File "<frozen importlib._bootstrap>", line 222, in _call_with_frames_removed
ImportError: DLL load failed: The specified module could not be found.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow.py", line 41, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 21, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "D:\Francesco\Anaconda\envs\DLC\lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 20, in swig_import_helper
    return importlib.import_module('_pywrap_tensorflow_internal')
  File "D:\Francesco\Anaconda\envs\DLC\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
ImportError: No module named '_pywrap_tensorflow_internal'


Failed to load the native TensorFlow runtime.

I installed the most recent version of win c++ distributable and checked that python is x64 as my system, but still does not work.

Edit:

list of packages in the env (currently i have cuda 8.0 and cudnn 5.1):

# Name                    Version                   Build  Channel
certifi                   2020.6.20          pyhd3eb1b0_3
numpy                     1.18.5                   pypi_0    pypi
pip                       10.0.1                   py35_0
protobuf                  3.15.8                   pypi_0    pypi
python                    3.5.6                he025d50_0
setuptools                40.2.0                   py35_0
six                       1.15.0                   pypi_0    pypi
tensorflow-gpu            1.1.0                    pypi_0    pypi
vc                        14.2                 h21ff451_1
vs2015_runtime            14.27.29016          h5e58377_2
werkzeug                  1.0.1                    pypi_0    pypi
wheel                     0.36.2             pyhd3eb1b0_0
wincertstore              0.2              py35hfebbdb8_0

re-edit:

today i tried again every possible combination of the software versions, don't know why but Cuda 8.0, Cudnn 5.1, Python 3.5.6 and Tensorflow 1.0.0 seems to work, but every other Cuda 8.0 combination that should work too do not. i am referring to this post: Which TensorFlow and CUDA version combinations are compatible? i'll stick to this combination not knowing why others don't work.

CoccoSyn
  • 1
  • 1

1 Answers1

0

TensorFlow-gpu 1.1.0 is tested and configured with CUDA 8.0 and cuDNN 5.1 as per te Tensorflow documentation.

Other combinations of CUDA, cuDNN and Tensorflow may break because of version incompatibility. It is highly recommended to use pre-reconfigured and pre-tested versions as mentioned in Tensorflow doc.