Clever Geek Handbook

GPU is not used for calculations even though tensor-stream-gpu is installed

The following software is installed on my computer: Anaconda (3), TensorFlow (GPU) and Keras. There are two Anaconda virtual environments, one with TensorFlow for Python 2.7 and one for 3.5, the GPU version installed according to TF instructions . (I had the TensorFlow processor version installed earlier in a separate environment, but I uninstalled it.)

When I run the following:

source activate tensorflow-gpu-3.5
python code.py

and check nvidia-smi

it only shows 3MiB memory usage in Python, so it looks like the GPU is not being used for calculations. ( code.py

- simple deep Q-learning algorithm implemented with Keras)

Any ideas what could possibly go wrong?

+3


source to share


3 answers


The reason my GPU was not working was because of a broken CuDNN installation, or rather the libraries and source are from different versions of CuDNN.



This has been corrected with the following tip.

0


source


A good way to debug these problems is to check which operations have been distributed to which devices.

You can check this by passing a config parameter to the session:

session = tf.Session(config=tf.ConfigProto(log_device_placement=True))


When you start the application, you will see some kind of output indicating which devices are in use.

You can find more information here: https://www.tensorflow.org/tutorials/using_gpu

+3


source


TensorFlow on Windows

It took me a few hours to fix problems with installing TensorFlow on windows, so here's a summary:

Check if TensorFlow-gpu is working or not (use this code):

with tf.device('/gpu:0'):
    a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
    b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
    c = tf.matmul(a, b)
with tf.Session() as sess:
    print(sess.run(c))

To check the list of available CPUs or GPUs (use this code):

from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())

if tf.test.gpu_device_name():
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))
else:
    print("Please install GPU version of TF")

Install Tensorflow GPU on Windows using CUDA and cuDNN

Guide overview

  • Install the Nvidia card on your computer along with the drivers.
  • Downloading and installing CUDA
  • Download and install "cuDNN"
  • Uninstalling Tensorflow, Installing Tensorflow GPU
  • Update% PATH% on the system
  • Check installation

Manual Full Details

Specify

  • You removed tensorflow and in order to get things done, you just installed tensorflow-gpu for it.
  • Remove tensorflow from pip and conda context depending on your project settings and only install tensorflow-gpu
  • After setting the PATH variable, be sure to log out or reboot the system.

Hope this is helpful :))

-1


source






More articles:


All Articles