That's why abstractions like PyTorch exist. You can write a few lines of Python and get good utilization of all those GPU cores.