Efficient way to run independent jobs on multiple GPU machine

Hi, does anyone have some ideas?