use example

[gpleiss]

use example

[taineleau-zz]

Hi, maybe you could try to put the model file under the project killer framework to use a more user-friendly APIs, such as reloading model and inference.

[taineleau-zz]

From the error message above, it seems that either the model is not converted to GPU mode (model = model.cuda()) or the images are not on GPU.

You must keep the model and the inputs on the same chip (i.e. all on CPUs or all on GPUs) to do the inference.

[chmaz]

Dear Tainelau,

Thanks a lot! Following your suggestion, this is how I corrected the problem:

outputs = model(Variable(image, volatile=True).cuda(async=True)).cpu()

Christophe