CODEWITHSUNDEEP
tensorflowkerasdeep-learninggpunvidia
jef
jef

Reputation: 4083

How to disable GPU in keras with tensorflow?

I want to compare processing time of my code with and without gpu. My backend of keras is Tensorflow. So it uses a GPU automatically. I use a model of keras/examples/mnist_mlp.py for comparing.

I checked the processing time like below. Then, how do I disable my GPU? Should ~/.keras/keras.json be modified?

$ time python mnist_mlp.py 
Test loss: 0.109761892007
Test accuracy: 0.9832
python mnist_mlp.py  38.22s user 3.18s system 162% cpu 25.543 total

Upvotes: 13

Views: 16673

Answers (2)

Jonas Byström
Jonas Byström

Reputation: 26139

$ CUDA_VISIBLE_DEVICES=-1 time python mnist_mlp.py

seems to be a) either the new way, or b) a way that works on both Windows and Linux.

Upvotes: 4

Nassim Ben
Nassim Ben

Reputation: 11553

Have you tried something like this? :

$ CUDA_VISIBLE_DEVICES='' time python mnist_mlp.py

CUDA_VISIBLE_DEVICES is usually used to hide some GPU's to cuda. Here you hide them all as you don't put any visible device.

Upvotes: 13

Related Questions