0
votes

All needed software are installed, Graphics card drivers for Gefore RTX 2060, CUDA 10.1, cuDNN 8.0.2.39, Anaconda3, TensorFlow 2.3.0. All according to Nvidias installation guide, making sure all versions work together.

However, I cannot find any GPU device from Jupyter Notebook. (Jupyter Notebook Code is provided down below)

Tensorflow 2.3.0 should automatically have GPU support, according to tensorflow.org. Which meens no need to install tensoflow-gpu. Right?

Your help is much appreciated. Thank you!


Underlying hardware and software:

windows 10 (64 bits)

Geforce RTX 2060 (driver version 442.23)

CUDA 10.1

cuDNN 8.0.2.39

Anaconda3

Tensorflow 2.3.0


import tensorflow as tf

import warnings

#check for GPU

if not tf.test.gpu_device_name():
   warnings.warn('No GPU found. Please ensure you have installed TensorFlow correctly')
else:
   print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))


#print TensorFlow version
print('TensorFlow Version: {}'.format(tf.__version__))

output:

<ipython-input-2-d8dd4f5b3689>:10: UserWarning: No GPU found. Please ensure you have installed TensorFlow correctly

TensorFlow Version: 2.3.0
1
Please include the full output of TensorFlow in your questionDr. Snoopy
The full output is already included. One line to check if GPU is found, and the other to print TensorFlow version.Alex
No, that is not the output I expect, it is the one that TensorFlow always outputs, including information about devices and if it found CUDA libraries.Dr. Snoopy
I am not sure what you mean. Are you talking about a specific method that I can run in Jupyter? If you provide it, I will run it and come back to you with the output.Alex
No, the normal tensorflow output will look like this: gist.github.com/mvaldenegro/69d497d9a4e01b744dc402cedc1d594a Note that it shows things like CUDA libraries being loaded, and the GPU being foundDr. Snoopy

1 Answers

0
votes

With CUDA 10.1 documentation says you should be using cuDNN v 7.6.5. Documentation is here