accelerate

fix: specify device for process_tensor in example usage

#3755

Merged