diff --git a/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_inner_ops.cc b/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_inner_ops.cc
index a65fb731380..f72af8ba6e6 100644
--- a/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_inner_ops.cc
+++ b/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_inner_ops.cc
@@ -236,7 +236,15 @@ REG_BPROP_BUILDER("ParallelResizeBilinear").SetUnusedInputs({i2}).SetBody(BODYFU
                       {"half_pixel_centers", MakeValue(false)}});
   return {dx, ib->ZerosLike(size)};
 });
-
+REG_BPROP_BUILDER("SiLU").SetUnusedInputs({i1}).SetBody(BODYFUNC(ib) {
+  auto x = ib->GetInput(kIndex0);
+  auto dout = ib->GetInput(kIndex2);
+  auto sigmoid_input = ib->Emit("Sigmoid", {x});
+  auto bc_dx = ib->Mul(x, dout);
+  auto bc_dy = ib->Mul(sigmoid_input, dout);
+  auto dx = ib->Emit("SigmoidGrad", {sigmoid_input, bc_dx});
+  return {ib->Add(dx, bc_dy)};
+});
 REG_BPROP_BUILDER("DynamicBroadcastTo").SetBody([](const BpropIRBuilder *ib) -> NodePtrList {
   auto x = ib->GetInput(kIndex0);
   auto shp = ib->GetInput(kIndex1);
diff --git a/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_math_ops.cc b/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_math_ops.cc
index b20f2e8e65e..21805a8af9e 100644
--- a/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_math_ops.cc
+++ b/mindspore/ccsrc/pipeline/pynative/grad/bprop_expander/grad_ops/grad_math_ops.cc
@@ -958,8 +958,13 @@ REG_BPROP_BUILDER("ReduceProd").SetUnusedInputs({i2}).SetBody(BODYFUNC(ib) {
                         {{"exclusive", MakeValue(true)}, {"reverse", MakeValue(true)}});
   auto y = ib->Reshape(ib->Mul(left, right), permuted_shape);
   auto out = ib->Mul(ib->Transpose(y, InvertPermutation(perm)), grad);
-  auto dx = ib->Reshape(out, input_shape);
-  return {dx, ib->ZerosLike(axis)};
+  auto x_dtype_id = ib->GetDtypeId(x);
+  if (x_dtype_id == kNumberTypeComplex64 || x_dtype_id == kNumberTypeComplex128) {
+    MS_EXCEPTION(TypeError) << "For 'ReduceProd', gradient not support for complex type currently.";
+  } else {
+    auto dx = ib->Reshape(out, input_shape);
+    return {dx, ib->ZerosLike(axis)};
+  }
 });
 
 REG_BPROP_BUILDER("ReduceMax").SetBody(BODYFUNC(ib) {
@@ -967,8 +972,13 @@ REG_BPROP_BUILDER("ReduceMax").SetBody(BODYFUNC(ib) {
   auto axis = ib->GetInput(kIndex1);
   auto out = ib->GetInput(kIndex2);
   auto dout = ib->GetInput(kIndex3);
-  auto dx = MinOrMaxGrad(ib, x, GetIntList(axis), out, dout);
-  return {dx, ib->ZerosLike(axis)};
+  auto x_dtype_id = ib->GetDtypeId(x);
+  if (x_dtype_id == kNumberTypeComplex64 || x_dtype_id == kNumberTypeComplex128) {
+    MS_EXCEPTION(TypeError) << "For 'ReduceMax', gradient not support for complex type currently.";
+  } else {
+    auto dx = MinOrMaxGrad(ib, x, GetIntList(axis), out, dout);
+    return {dx, ib->ZerosLike(axis)};
+  }
 });
 
 REG_BPROP_BUILDER("ReduceMin").SetBody(BODYFUNC(ib) {
@@ -976,8 +986,13 @@ REG_BPROP_BUILDER("ReduceMin").SetBody(BODYFUNC(ib) {
   auto axis = ib->GetInput(kIndex1);
   auto out = ib->GetInput(kIndex2);
   auto dout = ib->GetInput(kIndex3);
-  auto dx = MinOrMaxGrad(ib, x, GetIntList(axis), out, dout);
-  return {dx, ib->ZerosLike(axis)};
+  auto x_dtype_id = ib->GetDtypeId(x);
+  if (x_dtype_id == kNumberTypeComplex64 || x_dtype_id == kNumberTypeComplex128) {
+    MS_EXCEPTION(TypeError) << "For 'ReduceMin', gradient not support for complex type currently.";
+  } else {
+    auto dx = MinOrMaxGrad(ib, x, GetIntList(axis), out, dout);
+    return {dx, ib->ZerosLike(axis)};
+  }
 });
 
 REG_BPROP_BUILDER("ReduceMean").SetUnusedInputs({i0, i2}).SetBody(BODYFUNC(ib) {
@@ -1000,8 +1015,13 @@ REG_BPROP_BUILDER("ReduceMean").SetUnusedInputs({i0, i2}).SetBody(BODYFUNC(ib) {
     MS_EXCEPTION(ValueError) << "out shape size can not be 0";
   }
   auto div_shape = getSize(shape_x) / shape_out_sz;
-  auto dx = ib->RealDiv(grad, ib->Tensor(div_shape, ib->GetDtype(grad)));
-  return {dx, ib->ZerosLike(axis)};
+  auto x_dtype_id = ib->GetDtypeId(x);
+  if (x_dtype_id == kNumberTypeComplex64 || x_dtype_id == kNumberTypeComplex128) {
+    MS_EXCEPTION(TypeError) << "For 'ReduceMean', gradient not support for complex type currently.";
+  } else {
+    auto dx = ib->RealDiv(grad, ib->Tensor(div_shape, ib->GetDtype(grad)));
+    return {dx, ib->ZerosLike(axis)};
+  }
 });
 
 REG_BPROP_BUILDER("ArgMaxWithValue").SetBody(BODYFUNC(ib) {