diff --git a/mindspore/ccsrc/pre_activate/common/helper.cc b/mindspore/ccsrc/pre_activate/common/helper.cc
index 4cda390fbb2c8a1b908a4f7799b0f8d79319cd8a..9be537775e151af8d9bbac0e3eaa7a291f9db8a4 100644
--- a/mindspore/ccsrc/pre_activate/common/helper.cc
+++ b/mindspore/ccsrc/pre_activate/common/helper.cc
@@ -687,7 +687,7 @@ bool IsSameNode(const EquivPtr &equiv1, const EquivPtr &equiv2, const VarPtr &va
   MS_EXCEPTION_IF_NULL(equiv1_node);
   auto equiv2_node = GetAnfNodeByVar(equiv2, var_node);
   MS_EXCEPTION_IF_NULL(equiv2_node);
-  return equiv1_node == equiv2_node;
+  return *equiv1_node == *equiv2_node;
 }
 
 AnfNodePtr GetAnfNodeByVar(const EquivPtr &equiv, const VarPtr &var_node) {
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 f457148d513281b7f553dce20f50eaa3429534c7..8edf6d82f2400562be9407b9edd00b680df5afa7 100755
--- a/mindspore/ops/_grad/grad_math_ops.py
+++ b/mindspore/ops/_grad/grad_math_ops.py
@@ -191,8 +191,8 @@ 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