!1112 Fix weight decay bug for `SGD` optimizer

Merge pull request !1112 from seatea/fix-sgd-weight-decay-bug

mindspore/nn/optim/sgd.py

@@ -136,7 +136,6 @@ class SGD(Optimizer):
         params = self.parameters
         accum = self.accum
         stat = self.stat
-        gradients = self.decay_weight(gradients)
         gradients = self.scale_grad(gradients)
         lr = self.get_lr()
         if self.is_group_lr: