#calculate-loss-in-cpu

is it okay to calculate loss in cpu and backpropagate in torch?