Using GPU with gcnn_keras

[Rhys-McAlister]

Hi, I am using the torch backend with gcnn_keras but my gpu is not being used. If i called torch.cuda.isavaliable() I get true but I'm unsure what I need to do on the keras side to make this work any help would be appreciated.

[PatReis]

Hello, last time I checked gpu acceleration worked with kgcnn and torch.

A few things to check:

1. Does your pytorch installation works on GPU apart from keras_core.
2. Could you run a simple non-linear dense layer with keras_core on GPU and torch.

If all above works, then I would be interested what script are you trying to run on gpu.
I thought that all backend torch functions are supported by torch gpu but maybe I added something that isn't.
I would then have to start searching...
Note that with torch at the moment jagged input is not yet supported.