How to use importONNXfunction in GPU environment?

1 view (last 30 days)
I can't find the way to use importONNXfunction to use it at the gpu enviroment.
This is the code:
parallel.gpu.enableCUDAForwardCompatibility(true)
I = gpuArray(I);
params = importONNXFunction(modelfile,'UNet177Fcn');
result = UNet184Fcn(I,params,'Training',false);
when I change the input to the gpuarray, gpu works but the results going wrong.
there are so many placeholders to be replaced, so I think my model is more fit with the importONNXFunction.
So, is this the problem about the model? or just the importONNXfunction is doesn't fit with the gpu??

Answers (1)

Anshika Chaurasia
Anshika Chaurasia on 16 Jun 2021
Hi,
I tried to use importONNXFunction to import a onnx file and it gave me correct results on passing gpuArray input. There is no issue of importONNXFunction with gpu environment. May be there is some issue in the model you are using.
Hope it helps!

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!