In functions "targeted_detection" and "untargeted_detection", a SGD optimizer is initialized at the very beginning, but it seems that no "optimizer_s.step()" is run inside of the functions. May I know the reason why we need the "optimizer_s"? It seems that "loss.backward()" and "x_var.grad.data" can do the job without the "optimizer_s". Thanks.
Dear author,
In functions "targeted_detection" and "untargeted_detection", a SGD optimizer is initialized at the very beginning, but it seems that no "optimizer_s.step()" is run inside of the functions. May I know the reason why we need the "optimizer_s"? It seems that "loss.backward()" and "x_var.grad.data" can do the job without the "optimizer_s". Thanks.
Best regards, YH