diff --git a/torch_npu/csrc/aten/ops/GluGradKernelNpu.cpp b/torch_npu/csrc/aten/ops/GluGradKernelNpu.cpp
index 912c3d6e57e779a7ed3c9fb8958943a7ca4fb7cb..9ec04528db0c99168a1f47df549f7590686c0609 100644
--- a/torch_npu/csrc/aten/ops/GluGradKernelNpu.cpp
+++ b/torch_npu/csrc/aten/ops/GluGradKernelNpu.cpp
@@ -23,6 +23,12 @@ namespace at_npu {
 namespace native {
 
 at::Tensor& NPUNativeFunctions::glu_backward_out(const at::Tensor &grad_output, const at::Tensor &self, int64_t dim, at::Tensor &result) {
+  auto outputSize = input_same_output_size(self);
+  OpPreparation::CheckOut(
+      {grad_output, self},
+      result,
+      grad_output,
+      outputSize);
 
   TORCH_CHECK(self.dim() > 0, "glu does not support 0-dimensional Tensors");
   auto wrap_dim = at::maybe_wrap_dim(dim, self.dim());
diff --git a/torch_npu/csrc/aten/ops/GluKernelNpu.cpp b/torch_npu/csrc/aten/ops/GluKernelNpu.cpp
index c0514c39015fa0451f769f3f6335380f7fd5d09f..c444e51d7b395c6db86621ca2472ce9f1f3852fa 100644
--- a/torch_npu/csrc/aten/ops/GluKernelNpu.cpp
+++ b/torch_npu/csrc/aten/ops/GluKernelNpu.cpp
@@ -23,6 +23,11 @@ namespace at_npu {
 namespace native {
 
 at::Tensor& NPUNativeFunctions::glu_out(const at::Tensor& self, int64_t dim, at::Tensor& result) {
+  OpPreparation::CheckOut(
+      {self},
+      result,
+      self);
+
   TORCH_CHECK(self.dim() > 0, "glu does not support 0-dimensional at::Tensors");
   auto wrap_dim = at::maybe_wrap_dim(dim, self.dim());
   const int64_t nIn = self.size(wrap_dim);