AI Pool
Log in

Tensorflow 1.13 does not use GPU

I'm using keras 2.0 with tensorflow 1.13 backend. I had to upgrade my tensorflow. Solved all issues, but realized tensorflow is not using gpu. Also CUDA_VISIBLE_DEVICES = 0 is set. When I try to print all visible device in tensorflow, it prints CPU devices, but GPU is not there. What's wrong with it?

Keras: v2.1.*

Tensorflow: v1.13.1

CUDA: v10.1

Nvidia Driver: v418.39

tensorflowcudagputensorflow-gpu
3 votesNN138.00
1 Answers
LP190.00
3

The problem can be with tensorflow and tensorflow-gpu packages if you use pip. Try to remove tensorflow calling

pip uninstall tensorflow

After that check one more time you have tensorflow in your pc or not, calling

pip list | grep tensorflow

you can have another tensorflow after removing the previous one, that's why make sure you removed tensorflow package completely. After removing it, install tensorflow-gpu one more time (I recommend you to use  --upgrade flag as well to upgrade all dependencies).

pip install tensorflow-gpu[==version] [--upgrade] [--user]
Reply
Couldn't find what you were looking for?and we will find an expert to answer.
How helpful was this page?