diff --git a/mindspore/nn/optim/lamb.py b/mindspore/nn/optim/lamb.py
index b4d478f52ab38be605719c9f9dfa124dcd7b3240..a6a38f164a72216362ebb34268fd30264562c71b 100755
--- a/mindspore/nn/optim/lamb.py
+++ b/mindspore/nn/optim/lamb.py
@@ -180,7 +180,7 @@ class Lamb(Optimizer):
                  beta2=0.999,
                  eps=1e-6,
                  weight_decay=0.0,
-                 decay_filter=lambda x: 'LayerNorm' not in x.name and 'bias' not in x.name):
+                 decay_filter=lambda x: 'layernorm' not in x.name.lower() and 'bias' not in x.name.lower()):
 
         super(Lamb, self).__init__(start_learning_rate, params)
         if self.is_group:
diff --git a/mindspore/ops/_grad/grad_math_ops.py b/mindspore/ops/_grad/grad_math_ops.py
index c83d13a56d07bd97429fc9bb332eda0762475695..ee71979f28d5fdfb3010b6512598763a18d703b4 100755
--- a/mindspore/ops/_grad/grad_math_ops.py
+++ b/mindspore/ops/_grad/grad_math_ops.py
@@ -194,8 +194,8 @@ def get_bprop_mul(self):
     mul_func = P.Mul()
 
     def bprop(x, y, out, dout):
-        bc_dx = mul_func(dout, y)
-        bc_dy = mul_func(dout, x)
+        bc_dx = mul_func(y, dout)
+        bc_dy = mul_func(x, dout)
         return binop_grad_common(x, y, bc_dx, bc_dy)
 
     return bprop