0

I installed Nvidia drivers 9.1 on my Debian 9.7 (Dataproc) when I try to run TensorFlow 1.9 via this test script it fails:

Used this guide to install GPU Drivers: https://cloud.google.com/dataproc/docs/concepts/compute/gpus

Used pip install tensorflow-gpu

import tensorflow as tf

I get:

>>> import tensorflow as tf
Traceback (most recent call last):
  File "/opt/tony/TonY-samples/deps/tf-gpu/lib/python3.5/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/opt/tony/TonY-samples/deps/tf-gpu/lib/python3.5/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/opt/tony/TonY-samples/deps/tf-gpu/lib/python3.5/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 "/opt/tony/TonY-samples/deps/tf-gpu/lib/python3.5/imp.py", line 242, in load_module
    return load_dynamic(name, filename, file)
  File "/opt/tony/TonY-samples/deps/tf-gpu/lib/python3.5/imp.py", line 342, in load_dynamic
    return _load(spec)
ImportError: libcublas.so.9.0: cannot open shared object file: No such file or directory

CUDA version

cat /usr/lib/cuda/version.txt 
CUDA Version 9.1.85

TensorFlow (Tried with 1.9 and 1.12)

pip freeze
absl-py==0.7.0
astor==0.7.1
gast==0.2.2
grpcio==1.18.0
h5py==2.9.0
Keras-Applications==1.0.7
Keras-Preprocessing==1.0.9
Markdown==3.0.1
numpy==1.16.1
protobuf==3.6.1
six==1.12.0
tensorboard==1.12.2
tensorflow-gpu==1.12.0
termcolor==1.1.0
Werkzeug==0.14.1

Debian version

cat /etc/debian_version
9.7

I installed the drivers following this guide:

Followed this guide:

How to install CUDA 9.0 CUDNN 7.0 automatically without prompts? The apt-get doesn't have 9.0

  • Exactly how did you install CUDA on this machine? – Michael Hampton Feb 12 at 1:01
  • In the guide attached at the bottom – spicyramen Feb 12 at 3:33
  • OK, then how did you install tensorflow? – Michael Hampton Feb 12 at 3:47
  • pip install, seems to be that tensorflow gpu pipy package was compiled directly with CUDA 9.0 which force me to downgrade to CUDA 9.0 and CuDNN 7.X, I'm able to get the runnable files but require some type of manual interaction to Agree/accept terms. medium.com/@zhanwenchen/… – spicyramen Feb 12 at 4:33
0
./cuda-linux.9.0.176-22781540.run --noprompt

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.