I’m trying to use facenet with tensorflow gpu on ubuntu 18.4, my graphic card is Nvidia Quadro P2000. The problem is that, when i’m runnning the process i see my process on nvidia-smi but the percent of GPU usage still on 0% and sometime become 9 % but just a very few moment. And the time of detection still the same to CPU. Someone said its because of the acces to the data by the GPU but really don’t understand what it means.
for information i’m using cuda 10.0, because with cuda 10.1 i can’t show any process