Autokeras: Execute autokeras on GPU

[rafaelmarconiramos]

I try many different combinations of autokeras dependencies, but I never get to run autokeras on GPU.
Does somebody know what the problem is? Or the version of the libraries like pytorch, cudatoolkit, tensorflow-gpu ....

Could the project yield a docker container with autokeras and GPU support backend tensorflow?

I want to test autokeras, but it is impossible to execute my problem without GPU.

Thanks

[William8Work]

I was able to run the test examples from autokeras using GPU through tensorflow. Did you check if your environment has properly installed the CUDA drivers and tensorflow-gpu is loaded?

One way the check if the tensorflow is using GPU or not, run the following codes:
import tensorflow as tf
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))

You should see the output referencing the GPU if it's successful.

[rafaelmarconiramos]

Thanks. I got the tensorflow docker for GPU execution and installed the autokeras. Now it works.

[adrianog]

Can I ask you exactly what docker image you are running? Ty

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[crypdick]

@adrianog see https://autokeras.com/docker/

[cannin]

One way of checking for GPU on Tensorflow 2:

sess = tf.compat.v1.Session(config=tf.compat.v1.ConfigProto(log_device_placement=True))