diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 4e1b2ecf3671a179f2cd9783622fcb9af78fc8f2..75ae8be2cb74945c2a9783f887dd3f3f99a032a6 100755
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -27,7 +27,7 @@ repos:
         files: (api|backward)\.yaml$
 -   repo: local
     hooks:
-    -   id: clang-format-with-version-check
+    -   id: clang-format
         name: clang-format
         description: Format files with ClangFormat.
         entry: bash ./tools/codestyle/clang_format.hook -i
diff --git a/paddle/phi/kernels/gpu/abs_grad_kernel.cu b/paddle/phi/kernels/gpu/abs_grad_kernel.cu
index 1ce6a1638b1a04fa8e21adb386b1c92bd57296a2..2d96a7a88e33ea45ad138896017f6fe860619808 100644
--- a/paddle/phi/kernels/gpu/abs_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/abs_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/abs_grad_kernel.h"
+
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/abs_grad_kernel.h"
 #include "paddle/phi/kernels/impl/abs_grad_kernel_impl.h"
 
 using phi::dtype::complex;
diff --git a/paddle/phi/kernels/gpu/abs_kernel.cu b/paddle/phi/kernels/gpu/abs_kernel.cu
index 1364a1cd3fa2cbe861798c91bac97f8d7005b43c..d025f4b61e76315558768d2797cc7d91c6de098f 100644
--- a/paddle/phi/kernels/gpu/abs_kernel.cu
+++ b/paddle/phi/kernels/gpu/abs_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/abs_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/abs_kernel.h"
 #include "paddle/phi/kernels/funcs/complex_functors.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
 
diff --git a/paddle/phi/kernels/gpu/accuracy_kernel.cu b/paddle/phi/kernels/gpu/accuracy_kernel.cu
index d5b50800f53faaaf9fb05db0bfab54efe09de79e..771b189f04071228587866ca59274bc0e60b455e 100644
--- a/paddle/phi/kernels/gpu/accuracy_kernel.cu
+++ b/paddle/phi/kernels/gpu/accuracy_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/accuracy_kernel.h"
+
 #include <thrust/execution_policy.h>
 #include <thrust/reduce.h>
 
@@ -20,7 +22,6 @@
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/accuracy_kernel.h"
 
 namespace phi {
 using paddle::platform::PADDLE_CUDA_NUM_THREADS;
diff --git a/paddle/phi/kernels/gpu/activation_grad_kernel.cu b/paddle/phi/kernels/gpu/activation_grad_kernel.cu
index 8d7b3e8dde681df04e536d4f4b53c4216d043ce5..a127c9e913512a849593a8797f0da7f361be45b2 100644
--- a/paddle/phi/kernels/gpu/activation_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/activation_grad_kernel.cu
@@ -12,12 +12,13 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/activation_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_device_function.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/activation_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
 #include "paddle/phi/kernels/impl/activation_grad_impl.h"
 
diff --git a/paddle/phi/kernels/gpu/activation_kernel.cu b/paddle/phi/kernels/gpu/activation_kernel.cu
index 05ec5dfb840c08c6d54229b0db8d3387a066d34e..55df4268e0053481e4bda6617fc75647edc95530 100644
--- a/paddle/phi/kernels/gpu/activation_kernel.cu
+++ b/paddle/phi/kernels/gpu/activation_kernel.cu
@@ -12,12 +12,13 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/activation_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_device_function.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/activation_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
 #include "paddle/phi/kernels/impl/activation_grad_impl.h"
 #include "paddle/phi/kernels/impl/activation_impl.h"
diff --git a/paddle/phi/kernels/gpu/adadelta_kernel.cu b/paddle/phi/kernels/gpu/adadelta_kernel.cu
index 15c6ad18a9680c77bb7079b51ed7df7a03d17863..7516a277a746f12ec7c6326b7ebd8f64c789fc31 100644
--- a/paddle/phi/kernels/gpu/adadelta_kernel.cu
+++ b/paddle/phi/kernels/gpu/adadelta_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/adadelta_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/adadelta_kernel.h"
 #include "paddle/phi/kernels/impl/adadelta_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/adagrad_kernel.cu b/paddle/phi/kernels/gpu/adagrad_kernel.cu
index 381dde09be7d4fdeacecfe567b5aa0f92c0ca7d1..28db4d418b957868081e092c4fe340187bd7b28d 100644
--- a/paddle/phi/kernels/gpu/adagrad_kernel.cu
+++ b/paddle/phi/kernels/gpu/adagrad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/adagrad_kernel.h"
+
 #include "paddle/fluid/operators/math/selected_rows_functor.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/adagrad_kernel.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/impl/adagrad_kernel_impl.h"
 
diff --git a/paddle/phi/kernels/gpu/adam_kernel.cu b/paddle/phi/kernels/gpu/adam_kernel.cu
index edeeb64f5db5fa2dd542d18cd0bc931c60fb80b0..1322428270db5435b671c21b96f200e0806ea402 100644
--- a/paddle/phi/kernels/gpu/adam_kernel.cu
+++ b/paddle/phi/kernels/gpu/adam_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/adam_kernel.h"
+
 #include <math.h>  // for sqrt in CPU and CUDA
 
 #include <vector>
@@ -22,7 +24,6 @@
 #include "paddle/phi/common/amp_type_traits.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/adam_kernel.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/adam_functors.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
diff --git a/paddle/phi/kernels/gpu/adamax_kernel.cu b/paddle/phi/kernels/gpu/adamax_kernel.cu
index d9171c80fa9d824f3a31fae4fde5566b407d2e30..0817c531318c390ead201b1fda18da511ea5569f 100644
--- a/paddle/phi/kernels/gpu/adamax_kernel.cu
+++ b/paddle/phi/kernels/gpu/adamax_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/adamax_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/adamax_kernel.h"
 #include "paddle/phi/kernels/impl/adamax_kernel_impl.h"
 
 PD_REGISTER_KERNEL(adamax, GPU, ALL_LAYOUT, phi::AdamaxKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/adamw_kernel.cu b/paddle/phi/kernels/gpu/adamw_kernel.cu
index 7c00cd13d6b2604a3e6a05b6cd4b088422679ad9..cead67fd39a21f664b5013cba847bf84bdf26b7e 100644
--- a/paddle/phi/kernels/gpu/adamw_kernel.cu
+++ b/paddle/phi/kernels/gpu/adamw_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/adamw_kernel.h"
+
 #include <math.h>  // for sqrt in CPU and CUDA
 
 #include <vector>
@@ -22,7 +24,6 @@
 #include "paddle/phi/common/amp_type_traits.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/adamw_kernel.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/adam_functors.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
diff --git a/paddle/phi/kernels/gpu/add_n_kernel.cu b/paddle/phi/kernels/gpu/add_n_kernel.cu
index d109ec49cfdb99338715e2761b919d1ca0ac72a8..d165edaaed780ed218725e25ff3be2ce74fbf322 100644
--- a/paddle/phi/kernels/gpu/add_n_kernel.cu
+++ b/paddle/phi/kernels/gpu/add_n_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/add_n_kernel.h"
+
 #include "paddle/fluid/memory/malloc.h"
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/add_n_kernel.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 
diff --git a/paddle/phi/kernels/gpu/addmm_grad_kernel.cu b/paddle/phi/kernels/gpu/addmm_grad_kernel.cu
index be71f619aa17eacae683a637eeb8a70dae5cea01..65978da1374e4888afe8a7b408b0bb5a70d92b66 100644
--- a/paddle/phi/kernels/gpu/addmm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/addmm_grad_kernel.cu
@@ -12,9 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/addmm_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/addmm_grad_kernel.h"
 #include "paddle/phi/kernels/impl/addmm_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/addmm_kernel.cu b/paddle/phi/kernels/gpu/addmm_kernel.cu
index e30d664068e1761b61758bf9d2e7858b9cf17546..7b589ce20acca5c6cf51fd16ea223ef6b0d17466 100644
--- a/paddle/phi/kernels/gpu/addmm_kernel.cu
+++ b/paddle/phi/kernels/gpu/addmm_kernel.cu
@@ -12,9 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/addmm_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/addmm_kernel.h"
 #include "paddle/phi/kernels/impl/addmm_kernel_impl.h"
 
 PD_REGISTER_KERNEL(addmm, GPU, ALL_LAYOUT, phi::AddmmKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/allclose_kernel.cu b/paddle/phi/kernels/gpu/allclose_kernel.cu
index cfa809f60c2bc5e2faf0b43ea5847df2f258cfb0..8abc6b272c51158c355d91709abbfe1dedfa4f60 100644
--- a/paddle/phi/kernels/gpu/allclose_kernel.cu
+++ b/paddle/phi/kernels/gpu/allclose_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/allclose_kernel.h"
+
 #include "paddle/phi/core/enforce.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/allclose_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/arange_kernel.cu b/paddle/phi/kernels/gpu/arange_kernel.cu
index 99456147e5fedfb649e7874795c07e62c7412579..9ea0d7c5393c37cf51bd37be86a45c4b3432cc64 100644
--- a/paddle/phi/kernels/gpu/arange_kernel.cu
+++ b/paddle/phi/kernels/gpu/arange_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/arange_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/arange_kernel.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/range_function.h"
 
diff --git a/paddle/phi/kernels/gpu/arg_min_max_kernel.cu b/paddle/phi/kernels/gpu/arg_min_max_kernel.cu
index 101e01df52108ee0292f0a2159690b684ffa7807..b1a40f03e6194416347a984b96839031426f2fcb 100644
--- a/paddle/phi/kernels/gpu/arg_min_max_kernel.cu
+++ b/paddle/phi/kernels/gpu/arg_min_max_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/arg_min_max_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/arg_min_max_kernel.h"
 
 #if defined(__NVCC__) || defined(__HIPCC__)
 
diff --git a/paddle/phi/kernels/gpu/argsort_kernel.cu b/paddle/phi/kernels/gpu/argsort_kernel.cu
index 6e8d47d9a57f8c44a101c858d27df16e412de3b6..6a9c1e275998b882e3146e8969dbfe266aebc7a5 100644
--- a/paddle/phi/kernels/gpu/argsort_kernel.cu
+++ b/paddle/phi/kernels/gpu/argsort_kernel.cu
@@ -12,12 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/argsort_kernel.h"
+
 #include <thrust/copy.h>
 #include <thrust/execution_policy.h>
 #include <thrust/sequence.h>
 #include <thrust/sort.h>
-
-#include "paddle/phi/kernels/argsort_kernel.h"
 #ifdef __NVCC__
 #include "cub/cub.cuh"
 #endif
diff --git a/paddle/phi/kernels/gpu/auc_kernel.cu b/paddle/phi/kernels/gpu/auc_kernel.cu
index ef55fb9a5e3451b562bf36588836edf89314145a..5a1bb9874fe19a21c46994b06779e6ed2ef5dc4f 100644
--- a/paddle/phi/kernels/gpu/auc_kernel.cu
+++ b/paddle/phi/kernels/gpu/auc_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/auc_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/cpu/cpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/auc_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/bce_loss_grad_kernel.cu b/paddle/phi/kernels/gpu/bce_loss_grad_kernel.cu
index b9f1680726de8df94381019b3e8cf70464cc077c..94eabac4d1306ed170e54fc4945b4ca35481126c 100644
--- a/paddle/phi/kernels/gpu/bce_loss_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/bce_loss_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/bce_loss_grad_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/hostdevice.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/bce_loss_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/bce_loss_kernel.cu b/paddle/phi/kernels/gpu/bce_loss_kernel.cu
index bc2d278049ce9e6cb8fb07d897066b69063cd717..b190bce4742805bbfe18991e8b20c04a2339b719 100644
--- a/paddle/phi/kernels/gpu/bce_loss_kernel.cu
+++ b/paddle/phi/kernels/gpu/bce_loss_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/bce_loss_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/hostdevice.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/bce_loss_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
 
diff --git a/paddle/phi/kernels/gpu/bilinear_tensor_product_grad_kernel.cu b/paddle/phi/kernels/gpu/bilinear_tensor_product_grad_kernel.cu
index 6186e90d54ce139e09d9e66a4a3b45a113287a2f..907b84b0a3538c81a740398a8bc779915877136f 100644
--- a/paddle/phi/kernels/gpu/bilinear_tensor_product_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/bilinear_tensor_product_grad_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/bilinear_tensor_product_grad_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/bilinear_tensor_product_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(bilinear_tensor_product_grad,
diff --git a/paddle/phi/kernels/gpu/bilinear_tensor_product_kernel.cu b/paddle/phi/kernels/gpu/bilinear_tensor_product_kernel.cu
index b23e9ccfcc8eac7579cd5b6d47772e97c4de6523..201650f96491d9c1033f9c6ce98af32beac99e7c 100644
--- a/paddle/phi/kernels/gpu/bilinear_tensor_product_kernel.cu
+++ b/paddle/phi/kernels/gpu/bilinear_tensor_product_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/bilinear_tensor_product_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/bilinear_tensor_product_kernel_impl.h"
 
 PD_REGISTER_KERNEL(bilinear_tensor_product,
diff --git a/paddle/phi/kernels/gpu/bincount_kernel.cu b/paddle/phi/kernels/gpu/bincount_kernel.cu
index 257c708ee53143e1bfdc54a53b542028b9354f80..d6073193a15056dc22c72dff093852c186fcbd06 100644
--- a/paddle/phi/kernels/gpu/bincount_kernel.cu
+++ b/paddle/phi/kernels/gpu/bincount_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/bincount_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/bincount_kernel.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 
diff --git a/paddle/phi/kernels/gpu/broadcast_tensors_grad_kernel.cu b/paddle/phi/kernels/gpu/broadcast_tensors_grad_kernel.cu
index eb70ef9ee76f46b2ce611a3127ef02ec1e912000..02f06edd04f7da55d412e422ef8438edd7760a63 100644
--- a/paddle/phi/kernels/gpu/broadcast_tensors_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/broadcast_tensors_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/broadcast_tensors_grad_kernel.h"
+
 #include <vector>
 
 #include "paddle/fluid/framework/tensor_util.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/enforce.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/broadcast_tensors_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
 
diff --git a/paddle/phi/kernels/gpu/broadcast_tensors_kernel.cu b/paddle/phi/kernels/gpu/broadcast_tensors_kernel.cu
index 5c87c9fc907838c291a77d7838ed8d098d72348c..82dac4552a42febca6f8c159b87d53918c8d9f15 100644
--- a/paddle/phi/kernels/gpu/broadcast_tensors_kernel.cu
+++ b/paddle/phi/kernels/gpu/broadcast_tensors_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/broadcast_tensors_kernel.h"
+
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/broadcast_tensors_kernel.h"
 #include "paddle/phi/kernels/impl/broadcast_tensors_kernel_impl.h"
 
 PD_REGISTER_KERNEL(broadcast_tensors,
diff --git a/paddle/phi/kernels/gpu/cast_grad_kernel.cu b/paddle/phi/kernels/gpu/cast_grad_kernel.cu
index 0029e6e954e86418d3fd9ce26bef7d8014b26710..f4b610301583c515f72a29f367de2d9525cd10ad 100644
--- a/paddle/phi/kernels/gpu/cast_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/cast_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cast_grad_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/visit_type.h"
-#include "paddle/phi/kernels/cast_grad_kernel.h"
 #include "paddle/phi/kernels/gpu/cast_impl.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/cast_kernel.cu b/paddle/phi/kernels/gpu/cast_kernel.cu
index b2b42482ad357e048c597fa732ddeebef558b6d0..a879dc3bafd746450433f5a98281f5d8b35b801e 100644
--- a/paddle/phi/kernels/gpu/cast_kernel.cu
+++ b/paddle/phi/kernels/gpu/cast_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cast_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/visit_type.h"
-#include "paddle/phi/kernels/cast_kernel.h"
 #include "paddle/phi/kernels/gpu/cast_impl.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/channel_shuffle_grad_kernel.cu b/paddle/phi/kernels/gpu/channel_shuffle_grad_kernel.cu
index 5fc67daf44a4fabb40c6ef2bd31967928e9618b7..a9f751bfbc0a979958c57896dfbf1cd049cb386a 100644
--- a/paddle/phi/kernels/gpu/channel_shuffle_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/channel_shuffle_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/channel_shuffle_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/channel_shuffle_grad_kernel.h"
 #include "paddle/phi/kernels/impl/channel_shuffle_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(channel_shuffle_grad,
diff --git a/paddle/phi/kernels/gpu/channel_shuffle_kernel.cu b/paddle/phi/kernels/gpu/channel_shuffle_kernel.cu
index adecd2f9600361889593fb2ba96260944e00d6a7..6be097219524733c3f2c0e63cb2ed41af33377f2 100644
--- a/paddle/phi/kernels/gpu/channel_shuffle_kernel.cu
+++ b/paddle/phi/kernels/gpu/channel_shuffle_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/channel_shuffle_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/channel_shuffle_kernel.h"
 #include "paddle/phi/kernels/impl/channel_shuffle_kernel_impl.h"
 
 PD_REGISTER_KERNEL(channel_shuffle,
diff --git a/paddle/phi/kernels/gpu/cholesky_grad_kernel.cu b/paddle/phi/kernels/gpu/cholesky_grad_kernel.cu
index 0cff8de529428199a7207beada9500f4a7196240..9165e8ea4147ff02bfe8a84d8fc24e92a6826025 100644
--- a/paddle/phi/kernels/gpu/cholesky_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/cholesky_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cholesky_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/cholesky_grad_kernel.h"
 #include "paddle/phi/kernels/impl/cholesky_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/cholesky_kernel.cu b/paddle/phi/kernels/gpu/cholesky_kernel.cu
index ec9ac7545d23cfb8e12692f7d83e0ce16ec06908..66b3bf22b045b5d38d2636257b6d5081a06f6b7a 100644
--- a/paddle/phi/kernels/gpu/cholesky_kernel.cu
+++ b/paddle/phi/kernels/gpu/cholesky_kernel.cu
@@ -15,6 +15,8 @@ limitations under the License. */
 #ifndef PADDLE_WITH_HIP
 // HIP not support cusolver
 
+#include "paddle/phi/kernels/cholesky_kernel.h"
+
 #include <thrust/device_vector.h>
 
 #include <algorithm>
@@ -25,7 +27,6 @@ limitations under the License. */
 #include "paddle/phi/backends/dynload/cusolver.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/cholesky_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/clip_grad_kernel.cu b/paddle/phi/kernels/gpu/clip_grad_kernel.cu
index bab7dd41aee7ce181bc33256bc0c806101c0bbff..4566e8468ec1640dbc5b506b03199a27792b1373 100644
--- a/paddle/phi/kernels/gpu/clip_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/clip_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/clip_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/clip_grad_kernel.h"
 #include "paddle/phi/kernels/impl/clip_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(clip_grad,
diff --git a/paddle/phi/kernels/gpu/clip_kernel.cu b/paddle/phi/kernels/gpu/clip_kernel.cu
index 9295b8b37a01ffb53f9ee3dacd7815c5c7d4674b..9e0050db7fdbf178acb4fe5cf7174ebc951fc465 100644
--- a/paddle/phi/kernels/gpu/clip_kernel.cu
+++ b/paddle/phi/kernels/gpu/clip_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/clip_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/clip_kernel.h"
 #include "paddle/phi/kernels/impl/clip_kernel_impl.h"
 
 PD_REGISTER_KERNEL(clip,
diff --git a/paddle/phi/kernels/gpu/complex_grad_kernel.cu b/paddle/phi/kernels/gpu/complex_grad_kernel.cu
index b1a4c984eed7140619d32f89503240348f33da79..450b32291c4bc8b76f3fe6dcdad5e5e00dbad409 100644
--- a/paddle/phi/kernels/gpu/complex_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/complex_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/complex_grad_kernel.h"
+
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/complex_grad_kernel.h"
 #include "paddle/phi/kernels/impl/complex_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(imag_grad,
diff --git a/paddle/phi/kernels/gpu/complex_kernel.cu b/paddle/phi/kernels/gpu/complex_kernel.cu
index ae53aa95102281b6ee08c50231da91eb1b56753a..d0ee78202b0604de877edbc709fc565e25737a2b 100644
--- a/paddle/phi/kernels/gpu/complex_kernel.cu
+++ b/paddle/phi/kernels/gpu/complex_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/complex_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/complex_kernel.h"
 #include "paddle/phi/kernels/impl/complex_kernel_impl.h"
 
 // See Note [ Why still include the fluid headers? ]
diff --git a/paddle/phi/kernels/gpu/concat_grad_kernel.cu b/paddle/phi/kernels/gpu/concat_grad_kernel.cu
index 6b980c1a033358776ffe584ba18490ee838fe4fc..2445978daca462e877fdb1e6a0fd293674a0ffd8 100644
--- a/paddle/phi/kernels/gpu/concat_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/concat_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/concat_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/concat_grad_kernel.h"
 #include "paddle/phi/kernels/impl/concat_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(concat_grad,
diff --git a/paddle/phi/kernels/gpu/concat_kernel.cu b/paddle/phi/kernels/gpu/concat_kernel.cu
index 9582110c621cdeacb0c7009fba57c6dff1ffac26..accb1cc3d77e3ccd14b4d7808b781cf255eddd06 100644
--- a/paddle/phi/kernels/gpu/concat_kernel.cu
+++ b/paddle/phi/kernels/gpu/concat_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/concat_kernel.h"
+
 #include "paddle/fluid/operators/strided_memcpy.h"
 #include "paddle/fluid/platform/bfloat16.h"
 #include "paddle/fluid/platform/complex.h"
@@ -20,7 +22,6 @@
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/lod_utils.h"
-#include "paddle/phi/kernels/concat_kernel.h"
 #include "paddle/phi/kernels/funcs/concat_and_split_functor.h"
 #include "paddle/phi/kernels/funcs/concat_funcs.h"
 
diff --git a/paddle/phi/kernels/gpu/conv_grad_grad_kernel.cu b/paddle/phi/kernels/gpu/conv_grad_grad_kernel.cu
index f2669ebe04b2a8377b7f4445577d2582e54c773f..9d3e6da944a19fc51e5172f1cb936075416cd334 100644
--- a/paddle/phi/kernels/gpu/conv_grad_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/conv_grad_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_grad_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_grad_grad_kernel.h"
 #include "paddle/phi/kernels/impl/conv_grad_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/conv_grad_kernel.cu b/paddle/phi/kernels/gpu/conv_grad_kernel.cu
index 5fae327c2a93ff864faf6592a0934ad4abd372dc..e4b094314f8c1aff0057750e59a7e624a2e3fd03 100644
--- a/paddle/phi/kernels/gpu/conv_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/conv_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_grad_kernel.h"
 #include "paddle/phi/kernels/impl/conv_grad_kernel_impl.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/conv_kernel.cu b/paddle/phi/kernels/gpu/conv_kernel.cu
index 3aa406af4c2aec5306f4b42e28f99c7332227eff..e548b6fbf6d7b633c09d2dc869ec3f9b9e246938 100644
--- a/paddle/phi/kernels/gpu/conv_kernel.cu
+++ b/paddle/phi/kernels/gpu/conv_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_kernel.h"
 #include "paddle/phi/kernels/impl/conv_kernel_impl.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/conv_transpose_grad_kernel.cu b/paddle/phi/kernels/gpu/conv_transpose_grad_kernel.cu
index ee140a529b7fb4fc129012093414d6796d6503a7..0f7498a16d67af9673b7c333ab5c3aed04afe01c 100644
--- a/paddle/phi/kernels/gpu/conv_transpose_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/conv_transpose_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_transpose_grad_kernel.h"
+
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/ddim.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_transpose_grad_kernel.h"
 #include "paddle/phi/kernels/cpu/conv_util.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/gpu/depthwise_conv.h"
diff --git a/paddle/phi/kernels/gpu/conv_transpose_kernel.cu b/paddle/phi/kernels/gpu/conv_transpose_kernel.cu
index 4f8aae09a73855d0a760c9af81dd16e577cb21f9..94d0d536a85af18727c2180520fc63c194c63215 100644
--- a/paddle/phi/kernels/gpu/conv_transpose_kernel.cu
+++ b/paddle/phi/kernels/gpu/conv_transpose_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_transpose_kernel.h"
+
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/ddim.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_transpose_kernel.h"
 #include "paddle/phi/kernels/cpu/conv_util.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/gpu/depthwise_conv.h"
diff --git a/paddle/phi/kernels/gpu/copy_kernel.cu b/paddle/phi/kernels/gpu/copy_kernel.cu
index c917ce754890335f703bfb0f80677a9f49eb1a4e..16eff5b26e38a06950166079dc46d77c5a61a57c 100644
--- a/paddle/phi/kernels/gpu/copy_kernel.cu
+++ b/paddle/phi/kernels/gpu/copy_kernel.cu
@@ -12,11 +12,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/copy_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/data_type.h"
 #include "paddle/phi/core/compat/convert_utils.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/copy_kernel.h"
 
 // See Note [ Why still include the fluid headers? ]
 #include "paddle/fluid/memory/malloc.h"
diff --git a/paddle/phi/kernels/gpu/cross_grad_kernel.cu b/paddle/phi/kernels/gpu/cross_grad_kernel.cu
index 97d6d6849ae0048feef3f0f0422b73321e8c635c..ada78adb77fc97a62a179b0ebc60e0e89ce1e1b8 100644
--- a/paddle/phi/kernels/gpu/cross_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/cross_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cross_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/cross_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/cross_kernel.cu b/paddle/phi/kernels/gpu/cross_kernel.cu
index 4d4588f9b51c91cf1194b44e9bb1a2e955caeaa8..44173f4fbe62d98807e5671c98d27e38dbe5a2ac 100644
--- a/paddle/phi/kernels/gpu/cross_kernel.cu
+++ b/paddle/phi/kernels/gpu/cross_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cross_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/cross_kernel.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/cum_kernel.cu b/paddle/phi/kernels/gpu/cum_kernel.cu
index ad86fd9ba49dfc5a5a873cd24da0fceb965c6aa2..837914fb2db0621ba12fbc9442d9808513f8293a 100644
--- a/paddle/phi/kernels/gpu/cum_kernel.cu
+++ b/paddle/phi/kernels/gpu/cum_kernel.cu
@@ -12,12 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cum_kernel.h"
+
 #include <thrust/device_ptr.h>
 #include <thrust/device_vector.h>
 #include <thrust/reverse.h>
 #include <thrust/scan.h>
-
-#include "paddle/phi/kernels/cum_kernel.h"
 #ifdef __NVCC__
 #include <cub/cub.cuh>
 #endif
@@ -271,7 +271,6 @@ void ScanKernel(const Context& dev_ctx,
     return;
   }
 
-
   size_t height = 1;
   size_t width = 1;
   for (size_t i = 0; i <= axis; i++) {
@@ -391,9 +390,5 @@ PD_REGISTER_KERNEL(cumsum,
                    int,
                    int64_t) {}
 
-PD_REGISTER_KERNEL(logcumsumexp,
-                   GPU,
-                   ALL_LAYOUT,
-                   phi::LogcumsumexpKernel,
-                   float,
-                   double) {}
+PD_REGISTER_KERNEL(
+    logcumsumexp, GPU, ALL_LAYOUT, phi::LogcumsumexpKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/cumprod_grad_kernel.cu b/paddle/phi/kernels/gpu/cumprod_grad_kernel.cu
index bbae4fd130c7a69b0960e87546fadc7e3e5236b9..aee1aaa9555b62dafe9de764d9832dc41a395265 100644
--- a/paddle/phi/kernels/gpu/cumprod_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/cumprod_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cumprod_grad_kernel.h"
+
 #include <thrust/transform.h>
 
 #include "paddle/fluid/operators/math/inclusive_scan.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/cumprod_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/complex_functors.h"
 #include "paddle/phi/kernels/funcs/cumprod.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
diff --git a/paddle/phi/kernels/gpu/cumprod_kernel.cu b/paddle/phi/kernels/gpu/cumprod_kernel.cu
index 86aef50ac321cd9de812f6209ca444f6298e4391..1bbf8972a24798e5ddebf4dac2b3745eb1a2aee0 100644
--- a/paddle/phi/kernels/gpu/cumprod_kernel.cu
+++ b/paddle/phi/kernels/gpu/cumprod_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/cumprod_kernel.h"
+
 #include "paddle/fluid/operators/math/inclusive_scan.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/cumprod_kernel.h"
 #include "paddle/phi/kernels/funcs/complex_functors.h"
 #include "paddle/phi/kernels/funcs/cumprod.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
diff --git a/paddle/phi/kernels/gpu/deformable_conv_grad_kernel.cu b/paddle/phi/kernels/gpu/deformable_conv_grad_kernel.cu
index d80a4b8cc4cdec739d161eeb147f9ffb16fcb4bc..b46f1f4a3314dc4f6666b9e02fd3c7ea88e8bd50 100644
--- a/paddle/phi/kernels/gpu/deformable_conv_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/deformable_conv_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/deformable_conv_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/deformable_conv_grad_kernel.h"
 #include "paddle/phi/kernels/impl/deformable_conv_grad_kernel_impl.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/deformable_conv_kernel.cu b/paddle/phi/kernels/gpu/deformable_conv_kernel.cu
index 17a7b3265ca3edcc7f21b0fe7af09fca1b704073..2476dcbafb984856bee71f9a840ac6d45ba1b369 100644
--- a/paddle/phi/kernels/gpu/deformable_conv_kernel.cu
+++ b/paddle/phi/kernels/gpu/deformable_conv_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/deformable_conv_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/deformable_conv_kernel.h"
 #include "paddle/phi/kernels/impl/deformable_conv_kernel_impl.h"
 
 PD_REGISTER_KERNEL(deformable_conv,
diff --git a/paddle/phi/kernels/gpu/determinant_grad_kernel.cu b/paddle/phi/kernels/gpu/determinant_grad_kernel.cu
index 267a3b5e3fa5889c96eaff83f9049d7d1587770a..cce12a87fac72f5ac6edbbeb74de9fe3ae9ede09 100644
--- a/paddle/phi/kernels/gpu/determinant_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/determinant_grad_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/determinant_grad_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/determinant_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(determinant_grad,
diff --git a/paddle/phi/kernels/gpu/determinant_kernel.cu b/paddle/phi/kernels/gpu/determinant_kernel.cu
index b2dddf1fdb83f2fa0bbceabf7aa288bfcf615b7e..25184083873952638a1f84d8d4b66262363ca9c6 100644
--- a/paddle/phi/kernels/gpu/determinant_kernel.cu
+++ b/paddle/phi/kernels/gpu/determinant_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/determinant_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/determinant_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/diag_kernel.cu b/paddle/phi/kernels/gpu/diag_kernel.cu
index fd63084ecb35a5ead26e69856acb3a62b221024e..95d3d3365d91be61013e2016d06334f0498d866a 100644
--- a/paddle/phi/kernels/gpu/diag_kernel.cu
+++ b/paddle/phi/kernels/gpu/diag_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/diag_kernel.h"
+
 #include <algorithm>
 #include <tuple>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/diag_kernel.h"
 #include "paddle/phi/kernels/funcs/diag_functor.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 
diff --git a/paddle/phi/kernels/gpu/diagonal_grad_kernel.cu b/paddle/phi/kernels/gpu/diagonal_grad_kernel.cu
index 423093728e9d62386832d38a8db7caa5984e07d3..a5c0e05959842820eb0cf2612f79cd2932e1fba4 100644
--- a/paddle/phi/kernels/gpu/diagonal_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/diagonal_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/diagonal_grad_kernel.h"
+
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/diagonal_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/diagonal.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/diagonal_kernel.cu b/paddle/phi/kernels/gpu/diagonal_kernel.cu
index 58da29b2224a615234634f4e853089f5f51e2dcd..2e4ae5919971c43d674ac151b3327a55aa3b5c2f 100644
--- a/paddle/phi/kernels/gpu/diagonal_kernel.cu
+++ b/paddle/phi/kernels/gpu/diagonal_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/diagonal_kernel.h"
+
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/diagonal_kernel.h"
 #include "paddle/phi/kernels/funcs/diagonal.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/digamma_grad_kernel.cu b/paddle/phi/kernels/gpu/digamma_grad_kernel.cu
index 695227bba0f71d8c40e730a30998235c7c756442..5131c8f37f3bc263720f3eac0901dc5c89b74439 100644
--- a/paddle/phi/kernels/gpu/digamma_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/digamma_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/digamma_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/digamma_grad_kernel.h"
 #include "paddle/phi/kernels/impl/digamma_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/digamma_kernel.cu b/paddle/phi/kernels/gpu/digamma_kernel.cu
index 381c22a82e863d08d92b9fd5a9824fb9678ef2fa..31fc2295b02660db9347a32c1d1cdeccdcdb2549 100644
--- a/paddle/phi/kernels/gpu/digamma_kernel.cu
+++ b/paddle/phi/kernels/gpu/digamma_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/digamma_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/scalar.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/digamma_kernel.h"
 #include "paddle/phi/kernels/impl/digamma_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/dist_grad_kernel.cu b/paddle/phi/kernels/gpu/dist_grad_kernel.cu
index cdfec8d4afd713f6ae29a65d51b7db8c4fc21555..df422e8b2daf94c04e21ddbe9871a13783a601c2 100644
--- a/paddle/phi/kernels/gpu/dist_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/dist_grad_kernel.cu
@@ -12,9 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/dist_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/dist_grad_kernel.h"
 #include "paddle/phi/kernels/impl/dist_grad_kernel_impl.h"
 
 #ifdef PADDLE_WITH_HIP
diff --git a/paddle/phi/kernels/gpu/dist_kernel.cu b/paddle/phi/kernels/gpu/dist_kernel.cu
index d5ed1721612754c28154f35e6f46300235f98625..095110c25297833bbeada51542355c98c0a9316f 100644
--- a/paddle/phi/kernels/gpu/dist_kernel.cu
+++ b/paddle/phi/kernels/gpu/dist_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/dist_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/dist_kernel.h"
 #include "paddle/phi/kernels/impl/dist_kernel_impl.h"
 
 #ifdef PADDLE_WITH_HIP
diff --git a/paddle/phi/kernels/gpu/dot_grad_kernel.cu b/paddle/phi/kernels/gpu/dot_grad_kernel.cu
index c299e11b1f992a2578c031f6b98967d8e5b0ac7c..874d0f03b7dce3889cb313ef19d990e3e08f9066 100644
--- a/paddle/phi/kernels/gpu/dot_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/dot_grad_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/dot_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/dot_grad_kernel.h"
 #include "paddle/phi/kernels/impl/dot_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(dot_grad,
diff --git a/paddle/phi/kernels/gpu/dot_kernel.cu b/paddle/phi/kernels/gpu/dot_kernel.cu
index f947ef310eb223efb77176b388fbfc7fc08150c7..4442396f6c9dd752d044a1f540b673546944586c 100644
--- a/paddle/phi/kernels/gpu/dot_kernel.cu
+++ b/paddle/phi/kernels/gpu/dot_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/dot_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/dot_kernel.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 
 // See Note [ Why still include the fluid headers? ]
diff --git a/paddle/phi/kernels/gpu/dropout_grad_kernel.cu b/paddle/phi/kernels/gpu/dropout_grad_kernel.cu
index 1eea13a5a226b481cfdc80c4cfa6b9bcde24784b..841d98fbc003e547f5a11057b70b7dd6be6bc019 100644
--- a/paddle/phi/kernels/gpu/dropout_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/dropout_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/dropout_grad_kernel.h"
+
 #include "paddle/fluid/operators/dropout_impl.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/dropout_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/dropout_kernel.cu b/paddle/phi/kernels/gpu/dropout_kernel.cu
index 3811440be75113906881775971709087cdeecc9b..2fa3c7639e3960c8a9c3f9f4dc4f7b000c2ec2bb 100644
--- a/paddle/phi/kernels/gpu/dropout_kernel.cu
+++ b/paddle/phi/kernels/gpu/dropout_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/dropout_kernel.h"
+
 #include "paddle/fluid/operators/dropout_impl.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/dropout_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/eigh_grad_kernel.cu b/paddle/phi/kernels/gpu/eigh_grad_kernel.cu
index 51ff66b4a9ce7eb18e26014f5498be648f1c3f67..85bc5e6f76e54c4ee2870777f4e8161f541de369 100644
--- a/paddle/phi/kernels/gpu/eigh_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/eigh_grad_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/eigh_grad_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/complex_functors.h"
 #include "paddle/phi/kernels/impl/eigh_grad_kernel_impl.h"
 
diff --git a/paddle/phi/kernels/gpu/eigh_kernel.cu b/paddle/phi/kernels/gpu/eigh_kernel.cu
index 8c3ebf5f12b4b4d76ccc5c759a9c63dd76a2060e..3ffbb0b95b666523122265276c731c3252ccacc4 100644
--- a/paddle/phi/kernels/gpu/eigh_kernel.cu
+++ b/paddle/phi/kernels/gpu/eigh_kernel.cu
@@ -15,8 +15,9 @@
 #ifndef PADDLE_WITH_HIP
 // HIP not support cusolver
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/eigh_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/complex_functors.h"
 #include "paddle/phi/kernels/funcs/values_vectors_functor.h"
 
diff --git a/paddle/phi/kernels/gpu/einsum_kernel.cu b/paddle/phi/kernels/gpu/einsum_kernel.cu
index 088538908261f907af4c9ed6970c3a14fd0db0f3..d1f4c6590387a81464a3fdceec0442934e8b2940 100644
--- a/paddle/phi/kernels/gpu/einsum_kernel.cu
+++ b/paddle/phi/kernels/gpu/einsum_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/einsum_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/einsum_kernel.h"
 #include "paddle/phi/kernels/impl/einsum_impl.h"
 
 PD_REGISTER_KERNEL(einsum,
diff --git a/paddle/phi/kernels/gpu/elementwise_add_grad_kernel.cu b/paddle/phi/kernels/gpu/elementwise_add_grad_kernel.cu
index 4739d8cb3341f55c5fbccee03b5a7eded4886400..517fbcba158b8dfb65b69c22650d49a96e9389b1 100644
--- a/paddle/phi/kernels/gpu/elementwise_add_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/elementwise_add_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/elementwise_add_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
-#include "paddle/phi/kernels/elementwise_add_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
 #include "paddle/phi/kernels/gpu/elementwise_grad.h"
 #include "paddle/phi/kernels/impl/elementwise_grad_kernel_impl.h"
diff --git a/paddle/phi/kernels/gpu/elementwise_divide_grad_kernel.cu b/paddle/phi/kernels/gpu/elementwise_divide_grad_kernel.cu
index 5ef0752d6c8ba59c074b685bedfd374fe5c62f4c..57bf6da4060d34acdbf44ac80d4577e24247417d 100644
--- a/paddle/phi/kernels/gpu/elementwise_divide_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/elementwise_divide_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/elementwise_divide_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/elementwise_divide_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
 #include "paddle/phi/kernels/gpu/elementwise_grad.h"
 #include "paddle/phi/kernels/impl/elementwise_grad_kernel_impl.h"
diff --git a/paddle/phi/kernels/gpu/elementwise_grad_kernel.cu b/paddle/phi/kernels/gpu/elementwise_grad_kernel.cu
index 8dc0917fef575e25fe77cedb22f313ef6b116598..3e7430fd84eafbbea8c52384e10b79f6c028ab5a 100644
--- a/paddle/phi/kernels/gpu/elementwise_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/elementwise_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/elementwise_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
-#include "paddle/phi/kernels/elementwise_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
 #include "paddle/phi/kernels/gpu/elementwise_grad.h"
 #include "paddle/phi/kernels/impl/elementwise_grad_kernel_impl.h"
diff --git a/paddle/phi/kernels/gpu/elementwise_multiply_grad_kernel.cu b/paddle/phi/kernels/gpu/elementwise_multiply_grad_kernel.cu
index 0201854e533ddc7f027c0ceeb920c8eb03898788..3442d7f02853913a200c413fe1f7ebe0690e2627 100644
--- a/paddle/phi/kernels/gpu/elementwise_multiply_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/elementwise_multiply_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/elementwise_multiply_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/elementwise_multiply_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
 #include "paddle/phi/kernels/gpu/elementwise_grad.h"
 #include "paddle/phi/kernels/impl/elementwise_grad_kernel_impl.h"
diff --git a/paddle/phi/kernels/gpu/embedding_grad_kernel.cu b/paddle/phi/kernels/gpu/embedding_grad_kernel.cu
index d01674719edc7183d8be9ccdfc51d4a4bb81d0ee..f11fd0191b93514ffd0773d4c5c060e0797ed1be 100644
--- a/paddle/phi/kernels/gpu/embedding_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/embedding_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/embedding_grad_kernel.h"
+
 #include "paddle/fluid/framework/mixed_vector.h"
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/data_type.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/embedding_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/embedding_util.h"
 
diff --git a/paddle/phi/kernels/gpu/embedding_kernel.cu b/paddle/phi/kernels/gpu/embedding_kernel.cu
index c0516d00899f4aa8c1b7913201132ea146f95e68..90f3cc8d36032bab199236dd0d45cf2c4d82ac96 100644
--- a/paddle/phi/kernels/gpu/embedding_kernel.cu
+++ b/paddle/phi/kernels/gpu/embedding_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/embedding_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/data_type.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/embedding_kernel.h"
 #include "paddle/phi/kernels/funcs/eigen/eigen_function.h"
 #include "paddle/phi/kernels/funcs/embedding_util.h"
 
diff --git a/paddle/phi/kernels/gpu/erf_grad_kernel.cu b/paddle/phi/kernels/gpu/erf_grad_kernel.cu
index a06863b0a877714bc241d714429a8114f2b0d6f7..bbe7a20217f34cae437c94873123c22d1e665038 100644
--- a/paddle/phi/kernels/gpu/erf_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/erf_grad_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/erf_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/erf_grad_kernel.h"
 #include "paddle/phi/kernels/impl/erf_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(erf_grad,
diff --git a/paddle/phi/kernels/gpu/erf_kernel.cu b/paddle/phi/kernels/gpu/erf_kernel.cu
index 8e741be3345fc1d487e1a4148d96f0430f5978f0..cdd77424f60e1c8b4c765069edcdaae9d40dddb9 100644
--- a/paddle/phi/kernels/gpu/erf_kernel.cu
+++ b/paddle/phi/kernels/gpu/erf_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/erf_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/erf_kernel.h"
 #include "paddle/phi/kernels/impl/erf_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/erfinv_grad_kernel.cu b/paddle/phi/kernels/gpu/erfinv_grad_kernel.cu
index 078632fb4e647c656e0421e1d631b747f808f578..ce45d4e018e5a706a10922313683b31128c58221 100644
--- a/paddle/phi/kernels/gpu/erfinv_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/erfinv_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/erfinv_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/erfinv_grad_kernel.h"
 #include "paddle/phi/kernels/impl/erfinv_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/erfinv_kernel.cu b/paddle/phi/kernels/gpu/erfinv_kernel.cu
index 3296bb3dbb1673df7fd624a5a924ccb5d6c727f9..2316e960396fff4acc66e3f78eb58b47162630fa 100644
--- a/paddle/phi/kernels/gpu/erfinv_kernel.cu
+++ b/paddle/phi/kernels/gpu/erfinv_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/erfinv_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/erfinv_kernel.h"
 #include "paddle/phi/kernels/impl/erfinv_kernel_impl.h"
 
 PD_REGISTER_KERNEL(erfinv, GPU, ALL_LAYOUT, phi::ErfinvKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/expand_as_grad_kernel.cu b/paddle/phi/kernels/gpu/expand_as_grad_kernel.cu
index 387708af05bea70eda588a7be2181c247d673e7a..f289b8ce3bec800f6e32b45fbc80d6d49fd050eb 100644
--- a/paddle/phi/kernels/gpu/expand_as_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/expand_as_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/expand_as_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/expand_as_grad_kernel.h"
 #include "paddle/phi/kernels/impl/expand_as_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(expand_as_grad,
diff --git a/paddle/phi/kernels/gpu/expand_as_kernel.cu b/paddle/phi/kernels/gpu/expand_as_kernel.cu
index 68e683127baf24dd8219acc23e755e03ff75f72c..bd3f915c1a084ed1cee4e67a44e54c22cb548c69 100644
--- a/paddle/phi/kernels/gpu/expand_as_kernel.cu
+++ b/paddle/phi/kernels/gpu/expand_as_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/expand_as_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/expand_as_kernel.h"
 #include "paddle/phi/kernels/impl/expand_as_kernel_impl.h"
 
 PD_REGISTER_KERNEL(expand_as,
diff --git a/paddle/phi/kernels/gpu/expand_grad_kernel.cu b/paddle/phi/kernels/gpu/expand_grad_kernel.cu
index 8e2c3fde04a6a0038f457cfe74561719bba7f069..e8729942b6e6c9458c072f4baccea272b5512978 100644
--- a/paddle/phi/kernels/gpu/expand_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/expand_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/expand_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/expand_grad_kernel.h"
 #include "paddle/phi/kernels/impl/expand_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(expand_grad,
diff --git a/paddle/phi/kernels/gpu/expand_kernel.cu b/paddle/phi/kernels/gpu/expand_kernel.cu
index d4275804b3db8f4ae246379fe562d57598a2100d..27c4e82c6354440e892899f02e8e05171a504e1a 100644
--- a/paddle/phi/kernels/gpu/expand_kernel.cu
+++ b/paddle/phi/kernels/gpu/expand_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/expand_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/scalar.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/expand_kernel.h"
 #include "paddle/phi/kernels/impl/expand_kernel_impl.h"
 
 PD_REGISTER_KERNEL(expand,
diff --git a/paddle/phi/kernels/gpu/eye_kernel.cu b/paddle/phi/kernels/gpu/eye_kernel.cu
index d720feb9030c6d3fa5ab7978976ee35a5e28a6c7..5cd8e5caabe98089326a2dae614cd3b346d71016 100644
--- a/paddle/phi/kernels/gpu/eye_kernel.cu
+++ b/paddle/phi/kernels/gpu/eye_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/eye_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/eye_kernel.h"
 #include "paddle/phi/kernels/impl/eye_kernel_impl.h"
 
 PD_REGISTER_KERNEL(eye,
diff --git a/paddle/phi/kernels/gpu/flip_kernel.cu b/paddle/phi/kernels/gpu/flip_kernel.cu
index 519a57f32871d5b766daa1672292aedb3b860591..08ef6ddc5b2c22d40955244913ce74230bf35235 100644
--- a/paddle/phi/kernels/gpu/flip_kernel.cu
+++ b/paddle/phi/kernels/gpu/flip_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/flip_kernel.h"
+
 #include "paddle/fluid/memory/malloc.h"
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/place.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/flip_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/frobenius_norm_grad_kernel.cu b/paddle/phi/kernels/gpu/frobenius_norm_grad_kernel.cu
index 9011bb8c5d2ebabb086496502d6afe205d2009f7..d79f71014209c3f6c43788fa267e9c00d37a8941 100644
--- a/paddle/phi/kernels/gpu/frobenius_norm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/frobenius_norm_grad_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/frobenius_norm_grad_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/frobenius_norm_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(frobenius_norm_grad,
diff --git a/paddle/phi/kernels/gpu/frobenius_norm_kernel.cu b/paddle/phi/kernels/gpu/frobenius_norm_kernel.cu
index f41ccfd37735d9e9c9f0be5f1f9f44d1dcc72e64..a439711f5d04505c5dee34416e765779be9e1a99 100644
--- a/paddle/phi/kernels/gpu/frobenius_norm_kernel.cu
+++ b/paddle/phi/kernels/gpu/frobenius_norm_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/frobenius_norm_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/activation_functor.h"
 #include "paddle/phi/kernels/gpu/reduce.h"
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/full_kernel.cu b/paddle/phi/kernels/gpu/full_kernel.cu
index b823bb6aa6706604a6e810f60e98664a49c0397f..50e57a46317e35a7fc0cf1f27d496b88ffc846d3 100644
--- a/paddle/phi/kernels/gpu/full_kernel.cu
+++ b/paddle/phi/kernels/gpu/full_kernel.cu
@@ -12,9 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/full_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/full_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/gather_kernel.cu b/paddle/phi/kernels/gpu/gather_kernel.cu
index ffec56073e6c7178d87c521d8b123950603242ca..7e0c6cc168564e94c5af2e26a8f9ba4acc0594ed 100644
--- a/paddle/phi/kernels/gpu/gather_kernel.cu
+++ b/paddle/phi/kernels/gpu/gather_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/gather_kernel.h"
+
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
-#include "paddle/phi/kernels/gather_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/gather_nd_grad_kernel.cu b/paddle/phi/kernels/gpu/gather_nd_grad_kernel.cu
index 5273902804a200bdf36e8e36748639758145e742..a78dc717b046b61b54bc344f9f943b0faeb0ce50 100644
--- a/paddle/phi/kernels/gpu/gather_nd_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/gather_nd_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/gather_nd_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
-#include "paddle/phi/kernels/gather_nd_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/gather_nd_kernel.cu b/paddle/phi/kernels/gpu/gather_nd_kernel.cu
index 33745ef5f07e82387edfabf89d527b5b698fbabe..7b2412958902d3a98e9f612a06f4efdf1eb044bb 100644
--- a/paddle/phi/kernels/gpu/gather_nd_kernel.cu
+++ b/paddle/phi/kernels/gpu/gather_nd_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/gather_nd_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
-#include "paddle/phi/kernels/gather_nd_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/gather_tree_kernel.cu b/paddle/phi/kernels/gpu/gather_tree_kernel.cu
index 0172933d1be6021a32b2abb322236ae2446572b5..22b174b5f0bc21d6457ce31d203646f117252b99 100644
--- a/paddle/phi/kernels/gpu/gather_tree_kernel.cu
+++ b/paddle/phi/kernels/gpu/gather_tree_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/gather_tree_kernel.h"
+
 #include <algorithm>
 
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/gather_tree_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/gaussian_random_kernel.cu b/paddle/phi/kernels/gpu/gaussian_random_kernel.cu
index 0494c38b213a2dc2fd76993a0a19630a9a198bb6..f1edca70777fe71eb7074731df51522c6590d1fd 100644
--- a/paddle/phi/kernels/gpu/gaussian_random_kernel.cu
+++ b/paddle/phi/kernels/gpu/gaussian_random_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/gaussian_random_kernel.h"
+
 #include <thrust/random.h>
 
 #include "paddle/fluid/framework/generator.h"
@@ -21,7 +23,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/distribution_helper.h"
 #include "paddle/phi/kernels/funcs/index_impl.cu.h"
-#include "paddle/phi/kernels/gaussian_random_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/graph_reindex_kernel.cu b/paddle/phi/kernels/gpu/graph_reindex_kernel.cu
index c6be5231fe276180a58f52e8a951e9d3fc947772..046c210e186013692b89183d75b6c6461c084848 100644
--- a/paddle/phi/kernels/gpu/graph_reindex_kernel.cu
+++ b/paddle/phi/kernels/gpu/graph_reindex_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/graph_reindex_kernel.h"
+
 #include <thrust/copy.h>
 #include <thrust/device_vector.h>
 #include <thrust/reduce.h>
@@ -21,7 +23,6 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/graph_reindex_funcs.h"
-#include "paddle/phi/kernels/graph_reindex_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/graph_send_recv_grad_kernel.cu b/paddle/phi/kernels/gpu/graph_send_recv_grad_kernel.cu
index b00d9931646746d6c4743af3c454775e91b0ef4f..e78fb7892ed7d3da6c216453015a2d30b4c2e0ae 100644
--- a/paddle/phi/kernels/gpu/graph_send_recv_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/graph_send_recv_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/graph_send_recv_grad_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
@@ -19,7 +21,6 @@
 #include "paddle/phi/core/hostdevice.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/graph_send_recv_funcs.h"
-#include "paddle/phi/kernels/graph_send_recv_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/graph_send_recv_kernel.cu b/paddle/phi/kernels/gpu/graph_send_recv_kernel.cu
index 446a2361aedd852faeaa689884e3a2346500fe58..7ecf352ffe9966920a5fc8544633107013d39181 100644
--- a/paddle/phi/kernels/gpu/graph_send_recv_kernel.cu
+++ b/paddle/phi/kernels/gpu/graph_send_recv_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/graph_send_recv_kernel.h"
+
 #include <thrust/device_vector.h>
 #include <thrust/fill.h>
 
@@ -22,7 +24,6 @@
 #include "paddle/phi/core/hostdevice.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/graph_send_recv_funcs.h"
-#include "paddle/phi/kernels/graph_send_recv_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/grid_sample_grad_kernel.cu b/paddle/phi/kernels/gpu/grid_sample_grad_kernel.cu
index 40633fed348a53a98b788539f769c94f1755e08e..a258e041b17d572ab6c1e429aa7dda986fb94b2d 100644
--- a/paddle/phi/kernels/gpu/grid_sample_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/grid_sample_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/grid_sample_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_device_function.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/gpu/grid_sample_utils.h"
-#include "paddle/phi/kernels/grid_sample_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/grid_sample_kernel.cu b/paddle/phi/kernels/gpu/grid_sample_kernel.cu
index 4a5d567caa1ab94779f6ce64f90d143cc074f4dd..0bf0435240aae292ab056558090907a6287c4ba8 100644
--- a/paddle/phi/kernels/gpu/grid_sample_kernel.cu
+++ b/paddle/phi/kernels/gpu/grid_sample_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/grid_sample_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/grid_sample_utils.h"
-#include "paddle/phi/kernels/grid_sample_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/group_norm_grad_kernel.cu b/paddle/phi/kernels/gpu/group_norm_grad_kernel.cu
index 8af66fe0f29b6f1105f53524f1c0dc1a3310fe95..359dc8a0095f84528a79acad7893d11ecc8485b8 100644
--- a/paddle/phi/kernels/gpu/group_norm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/group_norm_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/group_norm_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/gpu/group_norm_utils.h"
-#include "paddle/phi/kernels/group_norm_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/group_norm_kernel.cu b/paddle/phi/kernels/gpu/group_norm_kernel.cu
index 127677233b80f6fe2846b4a968cd02f2929f1d89..012f224f04470faef4ddc9cfb52e71948f5d2090 100644
--- a/paddle/phi/kernels/gpu/group_norm_kernel.cu
+++ b/paddle/phi/kernels/gpu/group_norm_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/group_norm_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/gpu/group_norm_utils.h"
-#include "paddle/phi/kernels/group_norm_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/gumbel_softmax_grad_kernel.cu b/paddle/phi/kernels/gpu/gumbel_softmax_grad_kernel.cu
index 71d9859c8a3b3a80c4b08124a73d6d8f6546379a..2d7b4bdeb69167a8d08c08a60ae0dc9a3def4424 100644
--- a/paddle/phi/kernels/gpu/gumbel_softmax_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/gumbel_softmax_grad_kernel.cu
@@ -12,8 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gumbel_softmax_grad_kernel.h"
+
+#include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/gumbel_softmax_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(gumbel_softmax_grad,
diff --git a/paddle/phi/kernels/gpu/gumbel_softmax_kernel.cu b/paddle/phi/kernels/gpu/gumbel_softmax_kernel.cu
index d68c77de02fa4eff93b7a343b955b0ffcdf99ef5..33bf0eba380e446104059f35e37fb8ba556af16e 100644
--- a/paddle/phi/kernels/gpu/gumbel_softmax_kernel.cu
+++ b/paddle/phi/kernels/gpu/gumbel_softmax_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/gumbel_softmax_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/axis_utils.h"
-#include "paddle/phi/kernels/gumbel_softmax_kernel.h"
 #include "paddle/phi/kernels/impl/gumbel_softmax_kernel_impl.h"
 
 #if defined(__NVCC__) || defined(__HIPCC__)
diff --git a/paddle/phi/kernels/gpu/histogram_kernel.cu b/paddle/phi/kernels/gpu/histogram_kernel.cu
index 2950aef15ca935efe89eb27f7a391c70b7604195..cd89bf98a305bba23563f83c3110e1a162611afd 100644
--- a/paddle/phi/kernels/gpu/histogram_kernel.cu
+++ b/paddle/phi/kernels/gpu/histogram_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/histogram_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_launch_config.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/eigen/eigen_function.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/histogram_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/huber_loss_grad_kernel.cu b/paddle/phi/kernels/gpu/huber_loss_grad_kernel.cu
index 20cc2ed669adf91ee369b11ba957805839978581..9e169cbdcb7d0f88cb42e8d5201274d5d7c7170f 100644
--- a/paddle/phi/kernels/gpu/huber_loss_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/huber_loss_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/huber_loss_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/huber_loss_grad_kernel.h"
 #include "paddle/phi/kernels/impl/huber_loss_grad_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/huber_loss_kernel.cu b/paddle/phi/kernels/gpu/huber_loss_kernel.cu
index 26648a260b99ec6835ed3679e7c2c0550e8f6063..efb454832cd7f575cbc6cd7182b043dd597a5453 100644
--- a/paddle/phi/kernels/gpu/huber_loss_kernel.cu
+++ b/paddle/phi/kernels/gpu/huber_loss_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/huber_loss_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/huber_loss_kernel.h"
 #include "paddle/phi/kernels/impl/huber_loss_kernel_impl.h"
 
 PD_REGISTER_KERNEL(
diff --git a/paddle/phi/kernels/gpu/increment_kernel.cu b/paddle/phi/kernels/gpu/increment_kernel.cu
index 8e7cb6e8b3c05e220decea4666a3344e8d176ee9..b3c31271911489c94e895dbea786e4bf61f56bb4 100644
--- a/paddle/phi/kernels/gpu/increment_kernel.cu
+++ b/paddle/phi/kernels/gpu/increment_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/increment_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/increment_kernel_impl.h"
-#include "paddle/phi/kernels/increment_kernel.h"
 
 PD_REGISTER_KERNEL(increment,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/index_sample_grad_kernel.cu b/paddle/phi/kernels/gpu/index_sample_grad_kernel.cu
index b763f05531d00a78291fb9e11744760a18664daa..d2671dff7b0184f50345bdeeadb59d018a3badf0 100644
--- a/paddle/phi/kernels/gpu/index_sample_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/index_sample_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/index_sample_grad_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
@@ -21,7 +23,6 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/index_sample_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/index_sample_kernel.cu b/paddle/phi/kernels/gpu/index_sample_kernel.cu
index 702c955cd7e019c167801724bfd485515b4aa2dc..9b95d761fcbad475d142553741ffd658f0688884 100644
--- a/paddle/phi/kernels/gpu/index_sample_kernel.cu
+++ b/paddle/phi/kernels/gpu/index_sample_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/index_sample_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
@@ -20,7 +22,6 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/index_sample_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/index_select_grad_kernel.cu b/paddle/phi/kernels/gpu/index_select_grad_kernel.cu
index 75f74e4afdf1f232045b084c58c10105eb6b7bd7..8561744e87b1f85312867447c5a9672280744d6b 100644
--- a/paddle/phi/kernels/gpu/index_select_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/index_select_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/index_select_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_launch_config.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/utils/data_type.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/index_select_grad_kernel.h"
 
 DECLARE_bool(cudnn_deterministic);
 
diff --git a/paddle/phi/kernels/gpu/index_select_kernel.cu b/paddle/phi/kernels/gpu/index_select_kernel.cu
index d2a2bff075b195cba640d9e9ce2e0fd06fabba22..0a6ac69cef0981edebb2d273c04e103e9679e3ad 100644
--- a/paddle/phi/kernels/gpu/index_select_kernel.cu
+++ b/paddle/phi/kernels/gpu/index_select_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/index_select_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_launch_config.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/utils/data_type.h"
-#include "paddle/phi/kernels/index_select_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/instance_norm_grad_kernel.cu b/paddle/phi/kernels/gpu/instance_norm_grad_kernel.cu
index bdc81b59f143347997a4679fa05582533004f77d..d7ea2340afc7d36b85217a5bb4260e931888372a 100644
--- a/paddle/phi/kernels/gpu/instance_norm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/instance_norm_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/instance_norm_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/norm_utils.h"
 #include "paddle/phi/kernels/gpu/instance_norm_utils.h"
-#include "paddle/phi/kernels/instance_norm_grad_kernel.h"
 
 namespace phi {
 template <typename T, int BlockDim>
diff --git a/paddle/phi/kernels/gpu/instance_norm_kernel.cu b/paddle/phi/kernels/gpu/instance_norm_kernel.cu
index c7696c2dab8748d9d792e8b1e37aeba3d10c636c..b7292236898098a40fc4ee943abbe6da91818260 100644
--- a/paddle/phi/kernels/gpu/instance_norm_kernel.cu
+++ b/paddle/phi/kernels/gpu/instance_norm_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/instance_norm_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/norm_utils.h"
 #include "paddle/phi/kernels/gpu/instance_norm_utils.h"
-#include "paddle/phi/kernels/instance_norm_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/interpolate_grad_kernel.cu b/paddle/phi/kernels/gpu/interpolate_grad_kernel.cu
index 4b27e5dd359169a5029492c2f75f0a6e197d4bb0..175f09fccfa300fee21765d2fcb729aa251adf33 100644
--- a/paddle/phi/kernels/gpu/interpolate_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/interpolate_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/interpolate_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/fluid/platform/fast_divmod.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -21,7 +23,6 @@
 #include "paddle/phi/kernels/funcs/interpolate_function.h"
 #include "paddle/phi/kernels/funcs/math_cuda_utils.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/interpolate_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/interpolate_kernel.cu b/paddle/phi/kernels/gpu/interpolate_kernel.cu
index 108449c52ade3be68702085f89305e5559e8fc1a..7bc331c52a015d41118a87c46e79986c9d1b4426 100644
--- a/paddle/phi/kernels/gpu/interpolate_kernel.cu
+++ b/paddle/phi/kernels/gpu/interpolate_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/interpolate_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_device_function.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/fluid/platform/fast_divmod.h"
@@ -20,7 +22,6 @@
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/interpolate_function.h"
-#include "paddle/phi/kernels/interpolate_kernel.h"
 
 namespace phi {
 using paddle::platform::FastDivMod;
diff --git a/paddle/phi/kernels/gpu/isclose_kernel.cu b/paddle/phi/kernels/gpu/isclose_kernel.cu
index 20540521510b52b5598a25a2ef7eeec5ab5eb5cc..34774ec715c48de953945f94624c2c3cfe742d30 100644
--- a/paddle/phi/kernels/gpu/isclose_kernel.cu
+++ b/paddle/phi/kernels/gpu/isclose_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/isclose_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/isclose_kernel_impl.h"
-#include "paddle/phi/kernels/isclose_kernel.h"
 
 PD_REGISTER_KERNEL(
     isclose, GPU, ALL_LAYOUT, phi::IscloseKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/isfinite_kernel.cu b/paddle/phi/kernels/gpu/isfinite_kernel.cu
index 4b41ed1e55d39b537d89aa882ae10df1decbdd14..17ea25865170b4aabe871aaa1edd0f1cba30a86d 100644
--- a/paddle/phi/kernels/gpu/isfinite_kernel.cu
+++ b/paddle/phi/kernels/gpu/isfinite_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/isfinite_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/isfinite_kernel_impl.h"
-#include "paddle/phi/kernels/isfinite_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/kldiv_loss_grad_kernel.cu b/paddle/phi/kernels/gpu/kldiv_loss_grad_kernel.cu
index de3f6bc3f40c6fd1380018f6b59aa662fafffab2..8ca53f021f054278b813fedb69db31f4d2c5aaf6 100644
--- a/paddle/phi/kernels/gpu/kldiv_loss_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/kldiv_loss_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/kldiv_loss_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/kldiv_loss_grad_kernel_impl.h"
-#include "paddle/phi/kernels/kldiv_loss_grad_kernel.h"
 PD_REGISTER_KERNEL(
     kldiv_loss_grad, GPU, ALL_LAYOUT, phi::KLDivLossGradKernel, float, double) {
 }
diff --git a/paddle/phi/kernels/gpu/kldiv_loss_kernel.cu b/paddle/phi/kernels/gpu/kldiv_loss_kernel.cu
index adaf6963bb83ad51d0cef38a64fcbd88f08273df..9388ac7071c3197c2a8774ffd5f0d670cdf8a8b2 100644
--- a/paddle/phi/kernels/gpu/kldiv_loss_kernel.cu
+++ b/paddle/phi/kernels/gpu/kldiv_loss_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/kldiv_loss_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/kldiv_loss_kernel_impl.h"
-#include "paddle/phi/kernels/kldiv_loss_kernel.h"
 PD_REGISTER_KERNEL(
     kldiv_loss, GPU, ALL_LAYOUT, phi::KLDivLossKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/kron_grad_kernel.cu b/paddle/phi/kernels/gpu/kron_grad_kernel.cu
index 4f4e329ed84f327fc3b3b55e9ff72a070ebe538b..13ef2adaab3f32791e5c108b3f12b217e5dcea07 100644
--- a/paddle/phi/kernels/gpu/kron_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/kron_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/kron_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/kron_grad_kernel_impl.h"
-#include "paddle/phi/kernels/kron_grad_kernel.h"
 
 PD_REGISTER_KERNEL(kron_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/kron_kernel.cu b/paddle/phi/kernels/gpu/kron_kernel.cu
index 3d2b1573e890e9ca2deca8bd562f4c8a2d4fe192..a2124fd5af7d79cf6d1227a73105dd3e5b729547 100644
--- a/paddle/phi/kernels/gpu/kron_kernel.cu
+++ b/paddle/phi/kernels/gpu/kron_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/kron_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/kron_kernel_impl.h"
-#include "paddle/phi/kernels/kron_kernel.h"
 
 PD_REGISTER_KERNEL(kron,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/kthvalue_grad_kernel.cu b/paddle/phi/kernels/gpu/kthvalue_grad_kernel.cu
index 4b317e3b9424d6f36a233d1e07f4eab6a3c260c6..997cf7ccf57e7fd3e5fe3d906216920c6ba45173 100644
--- a/paddle/phi/kernels/gpu/kthvalue_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/kthvalue_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/kthvalue_grad_kernel.h"
+
 #include "paddle/fluid/operators/top_k_function_cuda.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/kthvalue_grad_kernel.h"
 
 namespace phi {
 static int getBlockSize(int col) {
diff --git a/paddle/phi/kernels/gpu/kthvalue_kernel.cu b/paddle/phi/kernels/gpu/kthvalue_kernel.cu
index bd2b16fb378e25525af1a7f1d63ef9e855da7dae..c677e26d0ae3076a0bc7b807ea9b87d266cfdba6 100644
--- a/paddle/phi/kernels/gpu/kthvalue_kernel.cu
+++ b/paddle/phi/kernels/gpu/kthvalue_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/kthvalue_kernel.h"
+
 #include "paddle/fluid/operators/top_k_function_cuda.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/eigen/eigen_function.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/kthvalue_kernel.h"
 
 namespace phi {
 inline int getBlockSize(int col) {
diff --git a/paddle/phi/kernels/gpu/label_smooth_grad_kernel.cu b/paddle/phi/kernels/gpu/label_smooth_grad_kernel.cu
index f30e8c3cdcf7a914b02284341c232de0599900cd..ddf7657fe528c5028bd956e9d5a81e4af66d292a 100644
--- a/paddle/phi/kernels/gpu/label_smooth_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/label_smooth_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/label_smooth_grad_kernel.h"
+
 #include "paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/label_smooth_grad_kernel.h"
 
 namespace phi {
 template <typename T>
diff --git a/paddle/phi/kernels/gpu/label_smooth_kernel.cu b/paddle/phi/kernels/gpu/label_smooth_kernel.cu
index 2bcb0ce5f3ada6f40432c7ec956f20c3cfa6bdb7..4a2f27d3018c47df3f6950453627de973c0d8d42 100644
--- a/paddle/phi/kernels/gpu/label_smooth_kernel.cu
+++ b/paddle/phi/kernels/gpu/label_smooth_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/label_smooth_kernel.h"
+
 #include <vector>
 
 #include "paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/label_smooth_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/layer_norm_grad_kernel.cu b/paddle/phi/kernels/gpu/layer_norm_grad_kernel.cu
index 5a399361aaac900adefcac54bbabacd3c358c25c..961937441e1cf9154eef19fc014acc74466ceeca 100644
--- a/paddle/phi/kernels/gpu/layer_norm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/layer_norm_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/layer_norm_grad_kernel.h"
+
 #include "paddle/fluid/operators/layer_norm_kernel.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/layer_norm_util.h"
-#include "paddle/phi/kernels/layer_norm_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/layer_norm_kernel.cu b/paddle/phi/kernels/gpu/layer_norm_kernel.cu
index 10aeba339cb042f7a78a2df820722e1de9a7848c..4922ba0cc2786bb4e57b1740a4370ff3f8d8cc52 100644
--- a/paddle/phi/kernels/gpu/layer_norm_kernel.cu
+++ b/paddle/phi/kernels/gpu/layer_norm_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/layer_norm_kernel.h"
+
 #include "paddle/fluid/operators/layer_norm_kernel.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/layer_norm_util.h"
-#include "paddle/phi/kernels/layer_norm_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/lerp_grad_kernel.cu b/paddle/phi/kernels/gpu/lerp_grad_kernel.cu
index 0a5ac99fa8e458cc7d786f1fbb00d8106032719c..39abe1e055d99bb4e0c838cbecb99f032ef688de 100644
--- a/paddle/phi/kernels/gpu/lerp_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/lerp_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/lerp_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/lerp_grad_kernel_impl.h"
-#include "paddle/phi/kernels/lerp_grad_kernel.h"
 
 PD_REGISTER_KERNEL(
     lerp_grad, GPU, ALL_LAYOUT, phi::LerpGradKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/lerp_kernel.cu b/paddle/phi/kernels/gpu/lerp_kernel.cu
index 96010aff4e70c6399032307ce7c48be7822ce12a..3f6862ff9795e2ae3bd78fdf0081edb88022b2e8 100644
--- a/paddle/phi/kernels/gpu/lerp_kernel.cu
+++ b/paddle/phi/kernels/gpu/lerp_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/lerp_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/lerp_kernel_impl.h"
-#include "paddle/phi/kernels/lerp_kernel.h"
 
 PD_REGISTER_KERNEL(lerp, GPU, ALL_LAYOUT, phi::LerpKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/lgamma_grad_kernel.cu b/paddle/phi/kernels/gpu/lgamma_grad_kernel.cu
index c7e82b8cd7e2d7baeae7899951dc8e93fd71bc59..3e4cd21a658f103aca9bc611a2d42518245e4401 100644
--- a/paddle/phi/kernels/gpu/lgamma_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/lgamma_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/lgamma_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/lgamma_grad_kernel_impl.h"
-#include "paddle/phi/kernels/lgamma_grad_kernel.h"
 PD_REGISTER_KERNEL(
     lgamma_grad, GPU, ALL_LAYOUT, phi::LgammaGradKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/lgamma_kernel.cu b/paddle/phi/kernels/gpu/lgamma_kernel.cu
index 050002f0555730e596f82dbd2d320a7dac0f29fc..e94d67f4ce324ad9d8237a377d70a920cdbd30af 100644
--- a/paddle/phi/kernels/gpu/lgamma_kernel.cu
+++ b/paddle/phi/kernels/gpu/lgamma_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/lgamma_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
-#include "paddle/phi/kernels/lgamma_kernel.h"
 
 namespace phi {
 template <typename T>
diff --git a/paddle/phi/kernels/gpu/linspace_kernel.cu b/paddle/phi/kernels/gpu/linspace_kernel.cu
index f16347cd3b6bfd20c48c68d39cade40b714b30e6..66a3f833d276a9c1644f513f92af809140aa48a5 100644
--- a/paddle/phi/kernels/gpu/linspace_kernel.cu
+++ b/paddle/phi/kernels/gpu/linspace_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/linspace_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/linspace_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/log_loss_grad_kernel.cu b/paddle/phi/kernels/gpu/log_loss_grad_kernel.cu
index bc14bd8f3c7df1d857345ce9bf467f5cfb567880..3bb256ad0326fdf1532fb3bf2162e0e64daba480 100644
--- a/paddle/phi/kernels/gpu/log_loss_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/log_loss_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/log_loss_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/log_loss_grad_kernel_impl.h"
-#include "paddle/phi/kernels/log_loss_grad_kernel.h"
 
 PD_REGISTER_KERNEL(
     log_loss_grad, GPU, ALL_LAYOUT, phi::LogLossGradKernel, float) {}
diff --git a/paddle/phi/kernels/gpu/log_loss_kernel.cu b/paddle/phi/kernels/gpu/log_loss_kernel.cu
index e7982b0b6fdf1bba73d481f1dd9a5a130b46fc00..0934520ea4ad14ce9450125b5fae7130140efa28 100644
--- a/paddle/phi/kernels/gpu/log_loss_kernel.cu
+++ b/paddle/phi/kernels/gpu/log_loss_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/log_loss_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/log_loss_kernel_impl.h"
-#include "paddle/phi/kernels/log_loss_kernel.h"
 
 PD_REGISTER_KERNEL(log_loss, GPU, ALL_LAYOUT, phi::LogLossKernel, float) {}
diff --git a/paddle/phi/kernels/gpu/log_softmax_grad_kernel.cu b/paddle/phi/kernels/gpu/log_softmax_grad_kernel.cu
index 78d1261df6f250b0bb65e40446b91c862808bd6a..f7b282536558db524c082de11c7ca92b2bd98edc 100644
--- a/paddle/phi/kernels/gpu/log_softmax_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/log_softmax_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/log_softmax_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpudnn/softmax_gpudnn.h"
-#include "paddle/phi/kernels/log_softmax_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/log_softmax_kernel.cu b/paddle/phi/kernels/gpu/log_softmax_kernel.cu
index b73bd6d6a9d9eb8ef63522d337341d1c6621ee7c..d7e34c6c14e7a49f50c016d888f6fb875dca0776 100644
--- a/paddle/phi/kernels/gpu/log_softmax_kernel.cu
+++ b/paddle/phi/kernels/gpu/log_softmax_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/log_softmax_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpudnn/softmax_gpudnn.h"
-#include "paddle/phi/kernels/log_softmax_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/logcumsumexp_grad_kernel.cu b/paddle/phi/kernels/gpu/logcumsumexp_grad_kernel.cu
index 43744210e32b7f2bb1ca66eaec512ab77d699027..9f4633a1e021ac77d9e47677b10fdc7ef9f8a9d2 100644
--- a/paddle/phi/kernels/gpu/logcumsumexp_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/logcumsumexp_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/logcumsumexp_grad_kernel.h"
+
 #include <limits>
 
 #include "paddle/phi/backends/cpu/cpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/logcumsumexp_grad_impl.h"
-#include "paddle/phi/kernels/logcumsumexp_grad_kernel.h"
 
 PD_REGISTER_KERNEL(logcumsumexp_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/logspace_kernel.cu b/paddle/phi/kernels/gpu/logspace_kernel.cu
index 673e8f0432015675a4e5f94f0633f551eb22d74a..95a196fe1b224110a1f76724b26005d4f6212a1c 100644
--- a/paddle/phi/kernels/gpu/logspace_kernel.cu
+++ b/paddle/phi/kernels/gpu/logspace_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/logspace_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/data_type_transform.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/logspace_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/logsumexp_grad_kernel.cu b/paddle/phi/kernels/gpu/logsumexp_grad_kernel.cu
index a5555bf7b5901a1220d8a28312c6f8b6ba8c1a8f..490b3e94045612cea551db515d273c7a0ef2c577 100644
--- a/paddle/phi/kernels/gpu/logsumexp_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/logsumexp_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/logsumexp_grad_kernel.h"
+
 #include "paddle/phi/backends/cpu/cpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/logsumexp_grad_kernel_impl.h"
-#include "paddle/phi/kernels/logsumexp_grad_kernel.h"
 
 PD_REGISTER_KERNEL(
     logsumexp_grad, GPU, ALL_LAYOUT, phi::LogsumexpGradKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/logsumexp_kernel.cu b/paddle/phi/kernels/gpu/logsumexp_kernel.cu
index c7c23fc307fe9be1d4292d0cd3307238aac82456..18249ff3bafb0d086134d106444473adb800bf5c 100644
--- a/paddle/phi/kernels/gpu/logsumexp_kernel.cu
+++ b/paddle/phi/kernels/gpu/logsumexp_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/logsumexp_kernel.h"
+
 #include "paddle/phi/backends/cpu/cpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/logsumexp_kernel_impl.h"
-#include "paddle/phi/kernels/logsumexp_kernel.h"
 
 PD_REGISTER_KERNEL(
     logsumexp, GPU, ALL_LAYOUT, phi::LogsumexpKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/masked_select_grad_kernel.cu b/paddle/phi/kernels/gpu/masked_select_grad_kernel.cu
index 171baab5513e444c91e6555a6c4cbee249859bf2..05f85ac1e613d770827b5f3053b92ebab48fb028 100644
--- a/paddle/phi/kernels/gpu/masked_select_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/masked_select_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/masked_select_grad_kernel.h"
+
 #include <thrust/device_ptr.h>
 #include <thrust/device_vector.h>
 #include <thrust/reverse.h>
@@ -19,7 +21,6 @@
 
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/select_impl.cu.h"
-#include "paddle/phi/kernels/masked_select_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/masked_select_kernel.cu b/paddle/phi/kernels/gpu/masked_select_kernel.cu
index b443ae6b8fb5e6c3bf5264a50d25205a419f22ad..632d79929cc3d0c707bbe6557f2257613c13daa3 100644
--- a/paddle/phi/kernels/gpu/masked_select_kernel.cu
+++ b/paddle/phi/kernels/gpu/masked_select_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/masked_select_kernel.h"
+
 #include <thrust/device_ptr.h>
 #include <thrust/device_vector.h>
 #include <thrust/reverse.h>
@@ -20,7 +22,6 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/select_impl.cu.h"
-#include "paddle/phi/kernels/masked_select_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/matmul_grad_kernel.cu b/paddle/phi/kernels/gpu/matmul_grad_kernel.cu
index b6c13360cd40423f4d9453e3478e676de80e9dd8..2753937eb7142c63d63067278bee04793e88bcce 100644
--- a/paddle/phi/kernels/gpu/matmul_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/matmul_grad_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/matmul_grad_kernel.h"
+
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/matmul_grad_kernel_impl.h"
-#include "paddle/phi/kernels/matmul_grad_kernel.h"
 
 PD_REGISTER_KERNEL(matmul_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/matmul_kernel.cu b/paddle/phi/kernels/gpu/matmul_kernel.cu
index 32d70ae0763f0fe4901204b8eace6c14c050571b..e5de7966c2ec40dc0eeeb960e979a195643d4c6e 100644
--- a/paddle/phi/kernels/gpu/matmul_kernel.cu
+++ b/paddle/phi/kernels/gpu/matmul_kernel.cu
@@ -12,11 +12,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/matmul_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/matmul_kernel_impl.h"
-#include "paddle/phi/kernels/matmul_kernel.h"
 
 PD_REGISTER_KERNEL(matmul,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/matrix_power_grad_kernel.cu b/paddle/phi/kernels/gpu/matrix_power_grad_kernel.cu
index 3739d7f2eedddfc4e7d82206e9fd2b8da572d983..25a9de8f8bed42121bcc6d52f4d2b2da565e4dfa 100644
--- a/paddle/phi/kernels/gpu/matrix_power_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/matrix_power_grad_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/matrix_power_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/matrix_power_grad_kernel_impl.h"
-#include "paddle/phi/kernels/matrix_power_grad_kernel.h"
 
 PD_REGISTER_KERNEL(matrix_power_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/matrix_power_kernel.cu b/paddle/phi/kernels/gpu/matrix_power_kernel.cu
index f474090f9db0971af43e7716dc36486d54412333..d7ae7d8a3f745c001c7ca82a1baffc8c1f6291c3 100644
--- a/paddle/phi/kernels/gpu/matrix_power_kernel.cu
+++ b/paddle/phi/kernels/gpu/matrix_power_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/matrix_power_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/matrix_power_kernel_impl.h"
-#include "paddle/phi/kernels/matrix_power_kernel.h"
 
 PD_REGISTER_KERNEL(
     matrix_power, GPU, ALL_LAYOUT, phi::MatrixPowerKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/matrix_rank_kernel.cu b/paddle/phi/kernels/gpu/matrix_rank_kernel.cu
index 8d69e6d896a529e8557fcc5dd30a5f02ab194309..9727e2cc114c54ded660fd16893089c1c3750e78 100644
--- a/paddle/phi/kernels/gpu/matrix_rank_kernel.cu
+++ b/paddle/phi/kernels/gpu/matrix_rank_kernel.cu
@@ -15,9 +15,10 @@
 #ifndef PADDLE_WITH_HIP
 // HIP not support cusolver
 
+#include "paddle/phi/kernels/matrix_rank_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/full_kernel.h"
-#include "paddle/phi/kernels/matrix_rank_kernel.h"
 #include "paddle/phi/kernels/matrix_rank_tol_kernel.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/matrix_rank_tol_kernel.cu b/paddle/phi/kernels/gpu/matrix_rank_tol_kernel.cu
index f3030d7f6cd7689efd6daa6c78b9237aedc863fc..c3aa64fe185d1d3e00e1247f553162134386f218 100644
--- a/paddle/phi/kernels/gpu/matrix_rank_tol_kernel.cu
+++ b/paddle/phi/kernels/gpu/matrix_rank_tol_kernel.cu
@@ -15,6 +15,8 @@
 #ifndef PADDLE_WITH_HIP
 // HIP not support cusolver
 
+#include "paddle/phi/kernels/matrix_rank_tol_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
@@ -27,7 +29,6 @@
 #include "paddle/phi/kernels/funcs/broadcast_function.h"
 #include "paddle/phi/kernels/funcs/compare_functors.h"
 #include "paddle/phi/kernels/impl/matrix_rank_kernel_impl.h"
-#include "paddle/phi/kernels/matrix_rank_tol_kernel.h"
 #include "paddle/phi/kernels/reduce_max_kernel.h"
 #include "paddle/phi/kernels/reduce_sum_kernel.h"
 
diff --git a/paddle/phi/kernels/gpu/mean_all_kernel.cu b/paddle/phi/kernels/gpu/mean_all_kernel.cu
index d87b738f4e717ff219405e0e3555766dbb23ec3e..da47bea57edb59b29204e93ed2283cb800f903e9 100644
--- a/paddle/phi/kernels/gpu/mean_all_kernel.cu
+++ b/paddle/phi/kernels/gpu/mean_all_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/mean_all_kernel.h"
+
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
-#include "paddle/phi/kernels/mean_all_kernel.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/mode_grad_kernel.cu b/paddle/phi/kernels/gpu/mode_grad_kernel.cu
index 77235c1da39cb1399bf1a81c4a98866fc438cdcc..43502621c2d3a878a144de1878aa09b8d64b6a47 100644
--- a/paddle/phi/kernels/gpu/mode_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/mode_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/mode_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/mode.h"
-#include "paddle/phi/kernels/mode_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/mode_kernel.cu b/paddle/phi/kernels/gpu/mode_kernel.cu
index ee255f10ebc5c473dfe2b1b0cf03adcff972de32..629b9722cd6bcfe12d0fb5a7e8be6439f5ea286f 100644
--- a/paddle/phi/kernels/gpu/mode_kernel.cu
+++ b/paddle/phi/kernels/gpu/mode_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/mode_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/mode.h"
-#include "paddle/phi/kernels/mode_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/momentum_kernel.cu b/paddle/phi/kernels/gpu/momentum_kernel.cu
index 5e00e074fe8f59f1454d5180493eac0d559bb8ce..5a4f5d33e6165370afb67960b8eb61200f034229 100644
--- a/paddle/phi/kernels/gpu/momentum_kernel.cu
+++ b/paddle/phi/kernels/gpu/momentum_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/momentum_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/momentum_kernel_impl.h"
-#include "paddle/phi/kernels/momentum_kernel.h"
 
 PD_REGISTER_KERNEL(momentum,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/multi_dot_grad_kernel.cu b/paddle/phi/kernels/gpu/multi_dot_grad_kernel.cu
index 61aeff9f3c70834b8bc22434f11baf5de642c093..9c665ed1233e006c2d8da2b049cf0543b694b659 100644
--- a/paddle/phi/kernels/gpu/multi_dot_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/multi_dot_grad_kernel.cu
@@ -12,11 +12,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/multi_dot_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/multi_dot_kernel_impl.h"
-#include "paddle/phi/kernels/multi_dot_grad_kernel.h"
 
 using float16 = phi::dtype::float16;
 
diff --git a/paddle/phi/kernels/gpu/multi_dot_kernel.cu b/paddle/phi/kernels/gpu/multi_dot_kernel.cu
index e890c03c34577a8e1ed52123459cc26de6c9d22c..be73e2d162a4bc841f5526c4b3478f0d1fcc4d53 100644
--- a/paddle/phi/kernels/gpu/multi_dot_kernel.cu
+++ b/paddle/phi/kernels/gpu/multi_dot_kernel.cu
@@ -12,11 +12,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/multi_dot_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/multi_dot_kernel_impl.h"
-#include "paddle/phi/kernels/multi_dot_kernel.h"
 
 using float16 = phi::dtype::float16;
 
diff --git a/paddle/phi/kernels/gpu/multiplex_grad_kernel.cu b/paddle/phi/kernels/gpu/multiplex_grad_kernel.cu
index 35258280f04da27835d6c55bc7123c8950a86a7a..21576ab608d269340322782c8113c6054c791e74 100644
--- a/paddle/phi/kernels/gpu/multiplex_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/multiplex_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/multiplex_grad_kernel.h"
+
 #include "paddle/phi/api/lib/utils/tensor_utils.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
-#include "paddle/phi/kernels/multiplex_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/multiplex_kernel.cu b/paddle/phi/kernels/gpu/multiplex_kernel.cu
index e1fbd7abdc4af42b28cb1a39e9d414f4b6e97ac7..743448a46866687cf2ac68be522a306281289252 100644
--- a/paddle/phi/kernels/gpu/multiplex_kernel.cu
+++ b/paddle/phi/kernels/gpu/multiplex_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/multiplex_kernel.h"
+
 #include "paddle/phi/api/lib/utils/tensor_utils.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/multiplex_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/mv_grad_kernel.cu b/paddle/phi/kernels/gpu/mv_grad_kernel.cu
index 58788492a741c21f78a3b6aa17d3c2cd590cab6d..5de66855f802e7a2ef41f794e79a6fed4101e938 100644
--- a/paddle/phi/kernels/gpu/mv_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/mv_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/mv_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/blas/blas.h"
-#include "paddle/phi/kernels/mv_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/mv_kernel.cu b/paddle/phi/kernels/gpu/mv_kernel.cu
index 82122723258de60665aa35b582a8587b6967ff83..fff548a9af3c5ca7fd589442a065f2fde05cea53 100644
--- a/paddle/phi/kernels/gpu/mv_kernel.cu
+++ b/paddle/phi/kernels/gpu/mv_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/mv_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/mv_kernel_impl.h"
-#include "paddle/phi/kernels/mv_kernel.h"
 
 PD_REGISTER_KERNEL(mv, GPU, ALL_LAYOUT, phi::MvKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/nanmedian_grad_kernel.cu b/paddle/phi/kernels/gpu/nanmedian_grad_kernel.cu
index d373e3bd9f3eeb726f8a8d0a9583e01dda285a91..b6333110485aff6d9cef572de0840a84e29c3684 100644
--- a/paddle/phi/kernels/gpu/nanmedian_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/nanmedian_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/nanmedian_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_launch_config.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/core/tensor_meta.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/impl/nanmedian_grad_kernel_impl.h"
-#include "paddle/phi/kernels/nanmedian_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/nanmedian_kernel.cu b/paddle/phi/kernels/gpu/nanmedian_kernel.cu
index 5ebf8637bfe22d5eeb7a45fd713da3377e0b2e48..f61e413c9887d1d31df6686da4d2e120fc400c5c 100644
--- a/paddle/phi/kernels/gpu/nanmedian_kernel.cu
+++ b/paddle/phi/kernels/gpu/nanmedian_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/nanmedian_kernel.h"
+
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/fluid/platform/device/gpu/gpu_launch_config.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/full_kernel.h"
 #include "paddle/phi/kernels/impl/nanmedian_kernel_impl.h"
-#include "paddle/phi/kernels/nanmedian_kernel.h"
 #include "paddle/phi/kernels/top_k_kernel.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/nll_loss_grad_kernel.cu b/paddle/phi/kernels/gpu/nll_loss_grad_kernel.cu
index 7b356826f5d8c7ee5586f8572302c3f948dd2d7a..7895983236f91d7b52556c8d293844f61be95a82 100644
--- a/paddle/phi/kernels/gpu/nll_loss_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/nll_loss_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/nll_loss_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/nll_loss.h"
-#include "paddle/phi/kernels/nll_loss_grad_kernel.h"
 
 namespace phi {
 template <typename T, typename Context>
diff --git a/paddle/phi/kernels/gpu/nll_loss_kernel.cu b/paddle/phi/kernels/gpu/nll_loss_kernel.cu
index bdb110fa92948611541a0ce8c86ff097622c52dc..1e80eb9bb460e19b137b4bdbbc8a6a3e256e051c 100644
--- a/paddle/phi/kernels/gpu/nll_loss_kernel.cu
+++ b/paddle/phi/kernels/gpu/nll_loss_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/nll_loss_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/nll_loss.h"
-#include "paddle/phi/kernels/nll_loss_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/norm_grad_kernel.cu b/paddle/phi/kernels/gpu/norm_grad_kernel.cu
index 388e7b889a1938d270a7ec2391f4a45618bfa920..bbbb6e9c018b61b590ba934b78ecffc7ee111a6e 100644
--- a/paddle/phi/kernels/gpu/norm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/norm_grad_kernel.cu
@@ -12,9 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include <algorithm>
-
 #include "paddle/phi/kernels/norm_grad_kernel.h"
+
+#include <algorithm>
 #ifdef __NVCC__
 #include "cub/cub.cuh"
 #endif
diff --git a/paddle/phi/kernels/gpu/norm_kernel.cu b/paddle/phi/kernels/gpu/norm_kernel.cu
index 2877069a226794a76f5bf1a4a1d9ad7270dd9436..bd9cffe79614b281dd3fa3d14501eec3922500c2 100644
--- a/paddle/phi/kernels/gpu/norm_kernel.cu
+++ b/paddle/phi/kernels/gpu/norm_kernel.cu
@@ -12,9 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include <algorithm>
-
 #include "paddle/phi/kernels/norm_kernel.h"
+
+#include <algorithm>
 #ifdef __NVCC__
 #include "cub/cub.cuh"
 #endif
diff --git a/paddle/phi/kernels/gpu/one_hot_kernel.cu b/paddle/phi/kernels/gpu/one_hot_kernel.cu
index adc87b049eefe810c41d570b83e5cda0ffbf90fb..2ae9e9333ecb568e73115570fa08db77e162d85b 100644
--- a/paddle/phi/kernels/gpu/one_hot_kernel.cu
+++ b/paddle/phi/kernels/gpu/one_hot_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/one_hot_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/one_hot_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/p_norm_grad_kernel.cu b/paddle/phi/kernels/gpu/p_norm_grad_kernel.cu
index 9305f30939fad53284bcca10474e2c102dcbdc46..fdfed25b3dda8fd03dcc1ab6a0fdccfc3dbb25a1 100644
--- a/paddle/phi/kernels/gpu/p_norm_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/p_norm_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/p_norm_grad_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/reduce_grad_functions.h"
-#include "paddle/phi/kernels/p_norm_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/p_norm_kernel.cu b/paddle/phi/kernels/gpu/p_norm_kernel.cu
index 12038fa22437ba64ebbc5195f79bad94db0e2178..80ef97d9cf88cda352a0b037cb449c0de6bc91c9 100644
--- a/paddle/phi/kernels/gpu/p_norm_kernel.cu
+++ b/paddle/phi/kernels/gpu/p_norm_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/p_norm_kernel.h"
+
 #include "paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h"
 #include "paddle/phi/common/amp_type_traits.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/gpu/reduce.h"
-#include "paddle/phi/kernels/p_norm_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/pad3d_grad_kernel.cu b/paddle/phi/kernels/gpu/pad3d_grad_kernel.cu
index 8832bf6a3a4cbc595367ecdbd1fd1d84ae915d5e..8f4af0a450890004bd09a650280583da8b961b95 100644
--- a/paddle/phi/kernels/gpu/pad3d_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/pad3d_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pad3d_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/pad3d_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/pad3d_kernel.cu b/paddle/phi/kernels/gpu/pad3d_kernel.cu
index eb8dfa527670883f88bbbab7c96225a1c927a314..d1b1d70667673e76de93711643490c5465d4bbb9 100644
--- a/paddle/phi/kernels/gpu/pad3d_kernel.cu
+++ b/paddle/phi/kernels/gpu/pad3d_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pad3d_kernel.h"
+
 #include <algorithm>
 
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/pad3d_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/pad_grad_kernel.cu b/paddle/phi/kernels/gpu/pad_grad_kernel.cu
index c5e2e077e4158e9fe36c3f5aef72fa92931e3494..a25472d122b837fcc3928af15e0678f0362abf0c 100644
--- a/paddle/phi/kernels/gpu/pad_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/pad_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pad_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pad_grad_kernel_impl.h"
-#include "paddle/phi/kernels/pad_grad_kernel.h"
 
 PD_REGISTER_KERNEL(pad_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/pad_kernel.cu b/paddle/phi/kernels/gpu/pad_kernel.cu
index 2b77a5f1aeb6cb3f24f274beb6939c480022fe49..cdf90513a3934ab14a63161329ad9e3ba6138d5c 100644
--- a/paddle/phi/kernels/gpu/pad_kernel.cu
+++ b/paddle/phi/kernels/gpu/pad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pad_kernel.h"
+
 #include "paddle/fluid/platform/complex.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pad_kernel_impl.h"
-#include "paddle/phi/kernels/pad_kernel.h"
 
 PD_REGISTER_KERNEL(pad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/pixel_shuffle_grad_kernel.cu b/paddle/phi/kernels/gpu/pixel_shuffle_grad_kernel.cu
index 6634d863fc14e82e406264e2d0f236428754de00..1414fb9df0b41ca5c857d04b2251dabe05c8a36f 100644
--- a/paddle/phi/kernels/gpu/pixel_shuffle_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/pixel_shuffle_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pixel_shuffle_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pixel_shuffle_grad_kernel_impl.h"
-#include "paddle/phi/kernels/pixel_shuffle_grad_kernel.h"
 
 PD_REGISTER_KERNEL(pixel_shuffle_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/pixel_shuffle_kernel.cu b/paddle/phi/kernels/gpu/pixel_shuffle_kernel.cu
index 8ceb1b70011613e280b1d2093b49602852ecbf29..e43d6f961236afcd51b29760c34d626464b3d0ca 100644
--- a/paddle/phi/kernels/gpu/pixel_shuffle_kernel.cu
+++ b/paddle/phi/kernels/gpu/pixel_shuffle_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pixel_shuffle_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pixel_shuffle_kernel_impl.h"
-#include "paddle/phi/kernels/pixel_shuffle_kernel.h"
 
 PD_REGISTER_KERNEL(
     pixel_shuffle, GPU, ALL_LAYOUT, phi::PixelShuffleKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/pixel_unshuffle_grad_kernel.cu b/paddle/phi/kernels/gpu/pixel_unshuffle_grad_kernel.cu
index f36c0e45174436c832aad852af2817af0c16a031..d7d2cde4ebade0b0acc1215ba691755ea76a37b0 100644
--- a/paddle/phi/kernels/gpu/pixel_unshuffle_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/pixel_unshuffle_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pixel_unshuffle_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pixel_unshuffle_grad_kernel_impl.h"
-#include "paddle/phi/kernels/pixel_unshuffle_grad_kernel.h"
 
 PD_REGISTER_KERNEL(pixel_unshuffle_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/pixel_unshuffle_kernel.cu b/paddle/phi/kernels/gpu/pixel_unshuffle_kernel.cu
index 54d29ab7b13ec920a36e5babe8e088bc10400124..fcc53cbee1ecb378b0a822a3528c6a4a75bdc91b 100644
--- a/paddle/phi/kernels/gpu/pixel_unshuffle_kernel.cu
+++ b/paddle/phi/kernels/gpu/pixel_unshuffle_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pixel_unshuffle_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pixel_unshuffle_kernel_impl.h"
-#include "paddle/phi/kernels/pixel_unshuffle_kernel.h"
 
 PD_REGISTER_KERNEL(pixel_unshuffle,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/pool_grad_kernel.cu b/paddle/phi/kernels/gpu/pool_grad_kernel.cu
index 832f9ad31188f03c2484113356b1481e46b9bc95..9cdf94b2be822f5f9fdcbc7d66c55c588d9a5ac4 100644
--- a/paddle/phi/kernels/gpu/pool_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/pool_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pool_grad_kernel.h"
+
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pool_grad_kernel_impl.h"
-#include "paddle/phi/kernels/pool_grad_kernel.h"
 
 PD_REGISTER_KERNEL(pool2d_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/pool_kernel.cu b/paddle/phi/kernels/gpu/pool_kernel.cu
index ac3718cfb80a2d5e0007795b046ad0c31f054bd0..81fce2060e47e206daf398863a3e276b1dd94b27 100644
--- a/paddle/phi/kernels/gpu/pool_kernel.cu
+++ b/paddle/phi/kernels/gpu/pool_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/pool_kernel.h"
+
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/pool_kernel_impl.h"
-#include "paddle/phi/kernels/pool_kernel.h"
 
 PD_REGISTER_KERNEL(pool2d,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/prelu_grad_kernel.cu b/paddle/phi/kernels/gpu/prelu_grad_kernel.cu
index 57d1838e90475e7094532b9d994a67cdc7eaa57d..c280873d8f768214807e9dec4e39541a9a529e73 100644
--- a/paddle/phi/kernels/gpu/prelu_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/prelu_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/prelu_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_meta.h"
 #include "paddle/phi/kernels/empty_kernel.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/gpu/prelu_funcs.h"
-#include "paddle/phi/kernels/prelu_grad_kernel.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/gpu/prelu_kernel.cu b/paddle/phi/kernels/gpu/prelu_kernel.cu
index ad87012485e76ba92f5b27261c0f46a7a8000b9b..c4730768982bb77b87bdfe1c11b08a74f72cface 100644
--- a/paddle/phi/kernels/gpu/prelu_kernel.cu
+++ b/paddle/phi/kernels/gpu/prelu_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/prelu_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/prelu_funcs.h"
-#include "paddle/phi/kernels/prelu_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/psroi_pool_kernel.cu b/paddle/phi/kernels/gpu/psroi_pool_kernel.cu
index d392ae7432fd60ffc6a46b0d11656d5b79570578..e0b17a55933b704264c18ff9978d4fa9e28cb531 100644
--- a/paddle/phi/kernels/gpu/psroi_pool_kernel.cu
+++ b/paddle/phi/kernels/gpu/psroi_pool_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/psroi_pool_kernel.h"
+
 #include <algorithm>
 #include <vector>
 
@@ -19,7 +21,6 @@
 #include "paddle/phi/common/place.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
-#include "paddle/phi/kernels/psroi_pool_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/put_along_axis_grad_kernel.cu b/paddle/phi/kernels/gpu/put_along_axis_grad_kernel.cu
index 209dd07d950c6c379ed22a7c21d7d4515c4f100d..f553da361f1fe825a17962ed2ac5c9463b509f6b 100644
--- a/paddle/phi/kernels/gpu/put_along_axis_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/put_along_axis_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/put_along_axis_grad_kernel.h"
+
 #include "paddle/fluid/framework/convert_utils.h"
 #include "paddle/fluid/operators/gather_scatter_kernel.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
-#include "paddle/phi/kernels/put_along_axis_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/put_along_axis_kernel.cu b/paddle/phi/kernels/gpu/put_along_axis_kernel.cu
index b52f2ae0ce199ff012478e6f73cc49a0764fbf54..d363c0c28364c065117fe53967234484871979af 100644
--- a/paddle/phi/kernels/gpu/put_along_axis_kernel.cu
+++ b/paddle/phi/kernels/gpu/put_along_axis_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/put_along_axis_kernel.h"
+
 #include "paddle/fluid/framework/convert_utils.h"
 #include "paddle/fluid/operators/gather_scatter_kernel.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
-#include "paddle/phi/kernels/put_along_axis_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/randint_kernel.cu b/paddle/phi/kernels/gpu/randint_kernel.cu
index 0882114e84672be5ba30555c4126bfc8a8d18e91..90eaea6a0868c1276687e9835b3133784806bfe1 100644
--- a/paddle/phi/kernels/gpu/randint_kernel.cu
+++ b/paddle/phi/kernels/gpu/randint_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/randint_kernel.h"
+
 #include <random>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/distribution_helper.h"
-#include "paddle/phi/kernels/randint_kernel.h"
 
 // See Note [ Why still include the fluid headers? ]
 #include "paddle/fluid/memory/memcpy.h"
diff --git a/paddle/phi/kernels/gpu/reduce_max_grad_kernel.cu b/paddle/phi/kernels/gpu/reduce_max_grad_kernel.cu
index 7ce58bf8b2b5f0fe32ca10701311743f56ab96f1..b4ff277b5026cec7d33ff697722dbb4888c601cf 100644
--- a/paddle/phi/kernels/gpu/reduce_max_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/reduce_max_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/reduce_max_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/reduce_max_grad_kernel_impl.h"
-#include "paddle/phi/kernels/reduce_max_grad_kernel.h"
 
 PD_REGISTER_KERNEL(max_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/reduce_mean_grad_kernel.cu b/paddle/phi/kernels/gpu/reduce_mean_grad_kernel.cu
index 57a86c63bfc7fead71982693008a4e2aa7696d3c..50564a339ddc05b48755a22ac2b9562eb94e41f7 100644
--- a/paddle/phi/kernels/gpu/reduce_mean_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/reduce_mean_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/reduce_mean_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/gpu/reduce_grad.h"
-#include "paddle/phi/kernels/reduce_mean_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/reduce_min_grad_kernel.cu b/paddle/phi/kernels/gpu/reduce_min_grad_kernel.cu
index 16914860491a3ff462fd3a56ffd41019334b5dff..ea1d377c4597640e1b0ee3db3e53d1a35ff5b46b 100644
--- a/paddle/phi/kernels/gpu/reduce_min_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/reduce_min_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/reduce_min_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/reduce_min_grad_kernel_impl.h"
-#include "paddle/phi/kernels/reduce_min_grad_kernel.h"
 
 PD_REGISTER_KERNEL(min_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/reduce_prod_grad_kernel.cu b/paddle/phi/kernels/gpu/reduce_prod_grad_kernel.cu
index 25f5ea33fbf7f43b578aa8413d90b8b1601bb850..08444cf95d6c6ab67118ff1a87ef1e3d9e7902e3 100644
--- a/paddle/phi/kernels/gpu/reduce_prod_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/reduce_prod_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/reduce_prod_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/reduce_prod_grad_kernel_impl.h"
-#include "paddle/phi/kernels/reduce_prod_grad_kernel.h"
 
 PD_REGISTER_KERNEL(prod_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/reduce_sum_grad_kernel.cu b/paddle/phi/kernels/gpu/reduce_sum_grad_kernel.cu
index f5d75b621c0be494b7ac751cc10ed1d2cc4fa039..8b111641cfa40f21dbf70f09984dfc493dfa0903 100644
--- a/paddle/phi/kernels/gpu/reduce_sum_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/reduce_sum_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/reduce_sum_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/gpu/reduce_grad.h"
-#include "paddle/phi/kernels/reduce_sum_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/rmsprop_kernel.cu b/paddle/phi/kernels/gpu/rmsprop_kernel.cu
index c49910e88b51aa1d44f66babf71f5fdf1bf81bd2..071c09ea675788bbae7741c5131fd33c6529684b 100644
--- a/paddle/phi/kernels/gpu/rmsprop_kernel.cu
+++ b/paddle/phi/kernels/gpu/rmsprop_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/rmsprop_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/rmsprop_kernel_impl.h"
-#include "paddle/phi/kernels/rmsprop_kernel.h"
 
 PD_REGISTER_KERNEL(
     rmsprop, GPU, ALL_LAYOUT, phi::RmspropDenseKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/roi_align_grad_kernel.cu b/paddle/phi/kernels/gpu/roi_align_grad_kernel.cu
index cfb9033bd9caae3cbbcf420f38c6ff86b6271a33..d173030a9c946a0220aa887476654e6b82f68b85 100644
--- a/paddle/phi/kernels/gpu/roi_align_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/roi_align_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/roi_align_grad_kernel.h"
+
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -20,7 +22,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/empty_kernel.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/roi_align_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/roi_align_kernel.cu b/paddle/phi/kernels/gpu/roi_align_kernel.cu
index 0bf96a729fca0593fabdf7b01a00b5235a6a27f0..a1824cc8fe47fadf50e1bfd6d5e8c57a6fab1095 100644
--- a/paddle/phi/kernels/gpu/roi_align_kernel.cu
+++ b/paddle/phi/kernels/gpu/roi_align_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/roi_align_kernel.h"
+
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/common/place.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/roi_align_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/roi_pool_grad_kernel.cu b/paddle/phi/kernels/gpu/roi_pool_grad_kernel.cu
index f66d6633b9ea07977f770bb1c6119ee23ea8f8d6..8b74e4bf2c68557198a67d86d39a0b589884efdd 100644
--- a/paddle/phi/kernels/gpu/roi_pool_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/roi_pool_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/roi_pool_grad_kernel.h"
+
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/common/place.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/roi_pool_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/roi_pool_kernel.cu b/paddle/phi/kernels/gpu/roi_pool_kernel.cu
index 4d3576f0c2f09277f8e119e9a549f76a2fd8561b..03c92e8727fcebcadec4c4b872d083a1ea9da311 100644
--- a/paddle/phi/kernels/gpu/roi_pool_kernel.cu
+++ b/paddle/phi/kernels/gpu/roi_pool_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/roi_pool_kernel.h"
+
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/common/place.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/roi_pool_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/roll_grad_kernel.cu b/paddle/phi/kernels/gpu/roll_grad_kernel.cu
index fc5d4ff538643c9159058e81696b4820ab579dcb..82e0fa72ab07607445e15b20822ba996a2269bd1 100644
--- a/paddle/phi/kernels/gpu/roll_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/roll_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/roll_grad_kernel.h"
+
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpu/roll_kernel_impl.h"
-#include "paddle/phi/kernels/roll_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/roll_kernel.cu b/paddle/phi/kernels/gpu/roll_kernel.cu
index 8b137e1a5aa0b2798326a81979e086b3f2a1aeb7..5d3584e4f44c1e68bc2ab4d6ff7b0ada39477a34 100644
--- a/paddle/phi/kernels/gpu/roll_kernel.cu
+++ b/paddle/phi/kernels/gpu/roll_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/roll_kernel.h"
+
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/utils/array.h"
 #include "paddle/phi/kernels/gpu/roll_kernel_impl.h"
-#include "paddle/phi/kernels/roll_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/rrelu_grad_kernel.cu b/paddle/phi/kernels/gpu/rrelu_grad_kernel.cu
index e2ebfc2ca027ddc9c2de8bcd9aa788ed53c3d8f2..fa9ef4503078559873a79126a13e91381ecb859e 100644
--- a/paddle/phi/kernels/gpu/rrelu_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/rrelu_grad_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/rrelu_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_meta.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/gpu/prelu_funcs.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
-#include "paddle/phi/kernels/rrelu_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/rrelu_kernel.cu b/paddle/phi/kernels/gpu/rrelu_kernel.cu
index 39582d5872a70ad2840cd0d135bfaf13fa416dad..e872cbf3cb6553ae58deb0fb234e5bc9e273fe68 100644
--- a/paddle/phi/kernels/gpu/rrelu_kernel.cu
+++ b/paddle/phi/kernels/gpu/rrelu_kernel.cu
@@ -12,11 +12,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/rrelu_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/distribution_helper.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
-#include "paddle/phi/kernels/rrelu_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/scale_kernel.cu b/paddle/phi/kernels/gpu/scale_kernel.cu
index 52b28bf37f02ebfbb3fb2b25d86ec09a55c03776..6f96a697b2f2db6c2097640f34c30142939f80e0 100644
--- a/paddle/phi/kernels/gpu/scale_kernel.cu
+++ b/paddle/phi/kernels/gpu/scale_kernel.cu
@@ -12,11 +12,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/scale_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
-#include "paddle/phi/kernels/scale_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/scatter_grad_kernel.cu b/paddle/phi/kernels/gpu/scatter_grad_kernel.cu
index 75506e2a0a17b269cb1327c01c1c4f3825eadb37..7f93fd0a9058c88127ab6443417b2cd3f8ef1fec 100644
--- a/paddle/phi/kernels/gpu/scatter_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/scatter_grad_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/scatter_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
-#include "paddle/phi/kernels/scatter_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/scatter_kernel.cu b/paddle/phi/kernels/gpu/scatter_kernel.cu
index 811eae1bc028ec5484d173d1b4373111546d73b4..af8919bec41cedba82c146924db18468ff2c4997 100644
--- a/paddle/phi/kernels/gpu/scatter_kernel.cu
+++ b/paddle/phi/kernels/gpu/scatter_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/scatter_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
-#include "paddle/phi/kernels/scatter_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/scatter_nd_add_grad_kernel.cu b/paddle/phi/kernels/gpu/scatter_nd_add_grad_kernel.cu
index 71924befe8cf9383c523eee059c3efa79dd6a262..66b373f3b2891e5f61610ff61bc4d4e5ff9647ea 100644
--- a/paddle/phi/kernels/gpu/scatter_nd_add_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/scatter_nd_add_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/scatter_nd_add_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
-#include "paddle/phi/kernels/scatter_nd_add_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/scatter_nd_add_kernel.cu b/paddle/phi/kernels/gpu/scatter_nd_add_kernel.cu
index 2ae8911fde510f36d192a6020bef4d37fe25611b..a7b8bebd38ce261a543651b3512e47569aa6366d 100644
--- a/paddle/phi/kernels/gpu/scatter_nd_add_kernel.cu
+++ b/paddle/phi/kernels/gpu/scatter_nd_add_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/scatter_nd_add_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
-#include "paddle/phi/kernels/scatter_nd_add_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/searchsorted_kernel.cu b/paddle/phi/kernels/gpu/searchsorted_kernel.cu
index 5976c14d9a9ea050ee389bc0befe79f83b02806e..4a2ce2241c22dc5c1cab391fe24a502ba845802b 100644
--- a/paddle/phi/kernels/gpu/searchsorted_kernel.cu
+++ b/paddle/phi/kernels/gpu/searchsorted_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/searchsorted_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/searchsorted_kernel_impl.h"
-#include "paddle/phi/kernels/searchsorted_kernel.h"
 
 PD_REGISTER_KERNEL(searchsorted,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/segment_pool_grad_kernel.cu b/paddle/phi/kernels/gpu/segment_pool_grad_kernel.cu
index 5f636ea7f9dd6c0bb66136eeab7d85d9d4acdc14..3cfbcc83cb13ee95bd00ec29050945b12940ae9f 100644
--- a/paddle/phi/kernels/gpu/segment_pool_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/segment_pool_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/segment_pool_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/segment_pool_grad_kernel_impl.h"
-#include "paddle/phi/kernels/segment_pool_grad_kernel.h"
 
 PD_REGISTER_KERNEL(segment_pool_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/segment_pool_kernel.cu b/paddle/phi/kernels/gpu/segment_pool_kernel.cu
index 4f24cf518d614020001fe97c419c298582b271fb..55ab822f5f54ebcb57939da99d748689c9e95a67 100644
--- a/paddle/phi/kernels/gpu/segment_pool_kernel.cu
+++ b/paddle/phi/kernels/gpu/segment_pool_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/segment_pool_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/segment_pool_kernel_impl.h"
-#include "paddle/phi/kernels/segment_pool_kernel.h"
 
 PD_REGISTER_KERNEL(segment_pool,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/selu_grad_kernel.cu b/paddle/phi/kernels/gpu/selu_grad_kernel.cu
index c6dffc33ae6c879a41a011ba49fe99cfd23020e1..0ed299413c1726f617dee9a8b5b4bf1d79d30efe 100644
--- a/paddle/phi/kernels/gpu/selu_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/selu_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/selu_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/selu_grad_kernel_impl.h"
-#include "paddle/phi/kernels/selu_grad_kernel.h"
 
 PD_REGISTER_KERNEL(
     selu_grad, GPU, ALL_LAYOUT, phi::SeluGradKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/selu_kernel.cu b/paddle/phi/kernels/gpu/selu_kernel.cu
index 57be1087b44c6be01eca0b3d7d45cd7f1a7779cb..99303d8c18a97893a939a8f358bac02603fae329 100644
--- a/paddle/phi/kernels/gpu/selu_kernel.cu
+++ b/paddle/phi/kernels/gpu/selu_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/selu_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/selu_kernel_impl.h"
-#include "paddle/phi/kernels/selu_kernel.h"
 
 PD_REGISTER_KERNEL(selu, GPU, ALL_LAYOUT, phi::SeluKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/set_value_grad_kernel.cu b/paddle/phi/kernels/gpu/set_value_grad_kernel.cu
index aafb374af63e292b94336777b54a9f1b909f284a..7eed96699e720870577c3d5246ce07c12c37335c 100644
--- a/paddle/phi/kernels/gpu/set_value_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/set_value_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/set_value_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/set_value_grad_kernel_impl.h"
-#include "paddle/phi/kernels/set_value_grad_kernel.h"
 
 PD_REGISTER_KERNEL(set_value_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/set_value_kernel.cu b/paddle/phi/kernels/gpu/set_value_kernel.cu
index b744cfc768e8d8b1a4c23b438ea2bb59700d5952..f788da010b6827d18ea455bad57d775da4049acf 100644
--- a/paddle/phi/kernels/gpu/set_value_kernel.cu
+++ b/paddle/phi/kernels/gpu/set_value_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/set_value_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/set_value_kernel_impl.h"
-#include "paddle/phi/kernels/set_value_kernel.h"
 
 PD_REGISTER_KERNEL(set_value,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/sgd_kernel.cu b/paddle/phi/kernels/gpu/sgd_kernel.cu
index 6d27843f138812a8c3727af6d88ccc422158f6a4..b2b73b2812c6075051b7149d5d698fb8f3767a82 100644
--- a/paddle/phi/kernels/gpu/sgd_kernel.cu
+++ b/paddle/phi/kernels/gpu/sgd_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/sgd_kernel.h"
+
 #include "paddle/fluid/framework/mixed_vector.h"
 #include "paddle/fluid/operators/amp/fp16_type_traits.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_helper.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/sgd_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/shard_index_kernel.cu b/paddle/phi/kernels/gpu/shard_index_kernel.cu
index b20c229a89fd10f9433c107bd8e96c7b12a00cef..d2497f56a0c76a5ca78f46163f608061de3240a7 100644
--- a/paddle/phi/kernels/gpu/shard_index_kernel.cu
+++ b/paddle/phi/kernels/gpu/shard_index_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/shard_index_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/shard_index_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_grad_kernel.cu b/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_grad_kernel.cu
index 8425f71cc265303a71a35174155d57a53e1caaff..f61cd2c39674ecdad5553e8472fb85da73ebd534 100644
--- a/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits.h"
 #include "paddle/phi/kernels/sigmoid_cross_entropy_with_logits_grad_kernel.h"
 
+#include "paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits.h"
+
 namespace phi {
 
 template <typename T>
diff --git a/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_kernel.cu b/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_kernel.cu
index 245ac95eeac677ffc9d56076e688ba8fd76f1537..b0e9efe5bbafe671ae6f594564260895a2015a45 100644
--- a/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_kernel.cu
+++ b/paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits.h"
 #include "paddle/phi/kernels/sigmoid_cross_entropy_with_logits_kernel.h"
 
+#include "paddle/phi/kernels/gpu/sigmoid_cross_entropy_with_logits.h"
+
 namespace phi {
 
 template <typename T>
diff --git a/paddle/phi/kernels/gpu/size_kernel.cu b/paddle/phi/kernels/gpu/size_kernel.cu
index 3ca1a1d6b760308ddbfa3ec2f49d305c6455d69a..fb6acd5599a8e5c9b81923c73dc045f0dc1a6f90 100644
--- a/paddle/phi/kernels/gpu/size_kernel.cu
+++ b/paddle/phi/kernels/gpu/size_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/size_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/size_kernel_impl.h"
-#include "paddle/phi/kernels/size_kernel.h"
 
 PD_REGISTER_KERNEL(size,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/softmax_grad_kernel.cu b/paddle/phi/kernels/gpu/softmax_grad_kernel.cu
index fe213e923a92f20e79cfe559c4b323819d3bf63b..04052e0dfc39a44a0f485557b9e8dc57b8794c38 100644
--- a/paddle/phi/kernels/gpu/softmax_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/softmax_grad_kernel.cu
@@ -12,12 +12,13 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/softmax_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/softmax_grad_kernel_impl.h"
-#include "paddle/phi/kernels/softmax_grad_kernel.h"
 
 PD_REGISTER_KERNEL(softmax_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/softmax_kernel.cu b/paddle/phi/kernels/gpu/softmax_kernel.cu
index 9415e6b2bad47746b499fc0aac989c3d54400696..4a02f438c7e7e4f0c0212ee613ce78a7fac20909 100644
--- a/paddle/phi/kernels/gpu/softmax_kernel.cu
+++ b/paddle/phi/kernels/gpu/softmax_kernel.cu
@@ -12,12 +12,13 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/softmax_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/softmax_kernel_impl.h"
-#include "paddle/phi/kernels/softmax_kernel.h"
 
 PD_REGISTER_KERNEL(softmax,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/split_kernel.cu b/paddle/phi/kernels/gpu/split_kernel.cu
index 9a854378fb54f9b3ad5e580241c75e22db473b7b..5f2741e53ddb99f0f749a6d95ce8b1d858b33444 100644
--- a/paddle/phi/kernels/gpu/split_kernel.cu
+++ b/paddle/phi/kernels/gpu/split_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/split_kernel.h"
+
 #include "paddle/fluid/operators/strided_memcpy.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/concat_and_split_functor.h"
-#include "paddle/phi/kernels/split_kernel.h"
 namespace phi {
 
 template <typename T, typename Context>
diff --git a/paddle/phi/kernels/gpu/squeeze_grad_kernel.cu b/paddle/phi/kernels/gpu/squeeze_grad_kernel.cu
index 8b2d00c5170d0dad91646f3ec14eaae95ec83807..c5a243f45bd9769b3f6a052248e979dad10e0a4f 100644
--- a/paddle/phi/kernels/gpu/squeeze_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/squeeze_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/squeeze_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/squeeze_grad_kernel_impl.h"
-#include "paddle/phi/kernels/squeeze_grad_kernel.h"
 
 PD_REGISTER_KERNEL(squeeze_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/squeeze_kernel.cu b/paddle/phi/kernels/gpu/squeeze_kernel.cu
index 6088e384c2e6b08d886c459f1b7d38016ebda8ca..ae15e210a02e72fd681d18fc6679f6e61c74e2bf 100644
--- a/paddle/phi/kernels/gpu/squeeze_kernel.cu
+++ b/paddle/phi/kernels/gpu/squeeze_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/squeeze_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/squeeze_kernel_impl.h"
-#include "paddle/phi/kernels/squeeze_kernel.h"
 
 PD_REGISTER_KERNEL(squeeze,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/stack_grad_kernel.cu b/paddle/phi/kernels/gpu/stack_grad_kernel.cu
index a24b48e0cf2c2d466e5e062f44a564a28363c058..97923eedcfc7110d45e19500495e67acd3b6b323 100644
--- a/paddle/phi/kernels/gpu/stack_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/stack_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/stack_grad_kernel.h"
+
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/stack_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/stack_kernel.cu b/paddle/phi/kernels/gpu/stack_kernel.cu
index 54bee95356ed32ecc8789768439d3c19433d9b40..22e8a3e48702dc6edef4da443cce72fda264573f 100644
--- a/paddle/phi/kernels/gpu/stack_kernel.cu
+++ b/paddle/phi/kernels/gpu/stack_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/stack_kernel.h"
+
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/stack_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/strided_slice_grad_kernel.cu b/paddle/phi/kernels/gpu/strided_slice_grad_kernel.cu
index cae05b67fefbe7a2a3acb62f37378d5d407925f4..90d9f1d9865773556df338a230097b977974de2e 100644
--- a/paddle/phi/kernels/gpu/strided_slice_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/strided_slice_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/strided_slice_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/strided_slice_grad_kernel_impl.h"
-#include "paddle/phi/kernels/strided_slice_grad_kernel.h"
 
 PD_REGISTER_KERNEL(strided_slice_raw_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/strided_slice_kernel.cu b/paddle/phi/kernels/gpu/strided_slice_kernel.cu
index fa7f5532f55fe433ce9f8586450271912ad886bc..716150ff47dea9b8f8166ac0fbd918ff2dbc8133 100644
--- a/paddle/phi/kernels/gpu/strided_slice_kernel.cu
+++ b/paddle/phi/kernels/gpu/strided_slice_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/strided_slice_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/complex.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/strided_slice_kernel_impl.h"
-#include "paddle/phi/kernels/strided_slice_kernel.h"
 
 PD_REGISTER_KERNEL(strided_slice_raw,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/take_along_axis_grad_kernel.cu b/paddle/phi/kernels/gpu/take_along_axis_grad_kernel.cu
index 61363ce8c49eb3c328c9296c0e72061d0d30e6dc..e09cfd370a4f0d87aaa081eaf50d519f536d2a72 100644
--- a/paddle/phi/kernels/gpu/take_along_axis_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/take_along_axis_grad_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/take_along_axis_grad_kernel.h"
+
 #include "paddle/fluid/framework/convert_utils.h"
 #include "paddle/fluid/operators/gather_scatter_kernel.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/take_along_axis_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/take_along_axis_kernel.cu b/paddle/phi/kernels/gpu/take_along_axis_kernel.cu
index cb567ef9915272d48323d3c587f0a99dff2f962f..9665a917d9dc4a5fa0d350f8fb635ecec79b7832 100644
--- a/paddle/phi/kernels/gpu/take_along_axis_kernel.cu
+++ b/paddle/phi/kernels/gpu/take_along_axis_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/take_along_axis_kernel.h"
+
 #include "paddle/fluid/framework/convert_utils.h"
 #include "paddle/fluid/operators/gather_scatter_kernel.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/take_along_axis_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/temporal_shift_grad_kernel.cu b/paddle/phi/kernels/gpu/temporal_shift_grad_kernel.cu
index 065b1726dce5b0f525e421eb5d2e094f1615badb..2bbfab9876bc81fc172ee99902fc98f4c55bee8e 100644
--- a/paddle/phi/kernels/gpu/temporal_shift_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/temporal_shift_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/temporal_shift_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/temporal_shift_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/temporal_shift_kernel.cu b/paddle/phi/kernels/gpu/temporal_shift_kernel.cu
index 34d80a1bc804b82bc93c77534ec7769b4b5406f4..990d882c2a5ead366b26182bef282968a3ac5e8b 100644
--- a/paddle/phi/kernels/gpu/temporal_shift_kernel.cu
+++ b/paddle/phi/kernels/gpu/temporal_shift_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/temporal_shift_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/layout.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/temporal_shift_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/tile_grad_kernel.cu b/paddle/phi/kernels/gpu/tile_grad_kernel.cu
index 32e78bbdb7cc7dab298454c1fc861cfdf9918cb3..c092609e623d3f4f3dc4b3d77b1c973e6ddfbcf3 100644
--- a/paddle/phi/kernels/gpu/tile_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/tile_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/tile_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/tile_grad_kernel_impl.h"
-#include "paddle/phi/kernels/tile_grad_kernel.h"
 
 PD_REGISTER_KERNEL(tile_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/tile_kernel.cu b/paddle/phi/kernels/gpu/tile_kernel.cu
index 40ee38f73b382a0d2ae1e857999c71629a862403..990877a8445cbf98d5b487a969a24059dbc24c84 100644
--- a/paddle/phi/kernels/gpu/tile_kernel.cu
+++ b/paddle/phi/kernels/gpu/tile_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/tile_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/tile_kernel_impl.h"
-#include "paddle/phi/kernels/tile_kernel.h"
 
 PD_REGISTER_KERNEL(tile,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/top_k_grad_kernel.cu b/paddle/phi/kernels/gpu/top_k_grad_kernel.cu
index 1db05ed18b0b9b77ce7125aafcb49d24ac9bb69f..8a8ea7ff12289fb48c6d02a3f746ac76ccb2a0d0 100644
--- a/paddle/phi/kernels/gpu/top_k_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/top_k_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/top_k_grad_kernel.h"
+
 #include "paddle/fluid/operators/top_k_function_cuda.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/top_k_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/top_k_kernel.cu b/paddle/phi/kernels/gpu/top_k_kernel.cu
index b160ad1d1c5dddd9a2c98d0df8178e23444469e0..e5038e0f3be7099d0d0ccbd6a225e12e6289769f 100644
--- a/paddle/phi/kernels/gpu/top_k_kernel.cu
+++ b/paddle/phi/kernels/gpu/top_k_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/top_k_kernel.h"
+
 #include "paddle/fluid/operators/top_k_function_cuda.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
-#include "paddle/phi/kernels/top_k_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/trace_grad_kernel.cu b/paddle/phi/kernels/gpu/trace_grad_kernel.cu
index 5b2575f48b87ea6b655d2189803ac2ed0f108b43..6692c1e19b033c3945387166f9954cc71fa6de32 100644
--- a/paddle/phi/kernels/gpu/trace_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/trace_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/trace_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/trace_grad_kernel_impl.h"
-#include "paddle/phi/kernels/trace_grad_kernel.h"
 
 PD_REGISTER_KERNEL(trace_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/trace_kernel.cu b/paddle/phi/kernels/gpu/trace_kernel.cu
index 7f22c14c6b609b6da8dd9e0f2b98cd044a7c2e52..4a749c5b3347da24c8aba35d33673801b4b7e407 100644
--- a/paddle/phi/kernels/gpu/trace_kernel.cu
+++ b/paddle/phi/kernels/gpu/trace_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/trace_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/diagonal.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
-#include "paddle/phi/kernels/trace_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/transpose_grad_kernel.cu b/paddle/phi/kernels/gpu/transpose_grad_kernel.cu
index 1439b2aa2ab86f201b2f238d215cef720c867fcd..e8ad456084ab147d9dea0f381bd5362be9bca633 100644
--- a/paddle/phi/kernels/gpu/transpose_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/transpose_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/transpose_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/transpose_grad_kernel_impl.h"
-#include "paddle/phi/kernels/transpose_grad_kernel.h"
 
 PD_REGISTER_KERNEL(transpose_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/transpose_kernel.cu b/paddle/phi/kernels/gpu/transpose_kernel.cu
index d7daaff2d413bec339163087f0daa55b79d9c946..62e29950e2d89ca3d27ddd148178dd6d69236501 100644
--- a/paddle/phi/kernels/gpu/transpose_kernel.cu
+++ b/paddle/phi/kernels/gpu/transpose_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/transpose_kernel.h"
+
 #include <vector>
 
 #include "paddle/fluid/framework/gpu_utils.h"
@@ -22,7 +24,6 @@
 #include "paddle/phi/common/bfloat16.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/transpose_grad_kernel_impl.h"
-#include "paddle/phi/kernels/transpose_kernel.h"
 
 namespace phi {
 template <typename T, typename Context>
diff --git a/paddle/phi/kernels/gpu/triangular_solve_kernel.cu b/paddle/phi/kernels/gpu/triangular_solve_kernel.cu
index 22629582bed8750f52a2f2789f3beb48f43dd11c..a48afeb2c796b70b24059a1cb87f1a1a3d665b35 100644
--- a/paddle/phi/kernels/gpu/triangular_solve_kernel.cu
+++ b/paddle/phi/kernels/gpu/triangular_solve_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/triangular_solve_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/ddim.h"
 #include "paddle/phi/core/kernel_registry.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/kernels/expand_kernel.h"
 #include "paddle/phi/kernels/funcs/blas/blas.h"
 #include "paddle/phi/kernels/funcs/common_shape.h"
-#include "paddle/phi/kernels/triangular_solve_kernel.h"
 
 // See Note [ Why still include the fluid headers? ]
 #include "paddle/fluid/memory/allocation/allocator.h"
diff --git a/paddle/phi/kernels/gpu/tril_indices_kernel.cu b/paddle/phi/kernels/gpu/tril_indices_kernel.cu
index 7068ab1775f1615e4be078a8df623d3f9255557a..be83f28451166b1193098ab746abc2a28cf4c028 100644
--- a/paddle/phi/kernels/gpu/tril_indices_kernel.cu
+++ b/paddle/phi/kernels/gpu/tril_indices_kernel.cu
@@ -12,12 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/tril_indices_kernel.h"
+
 #include <algorithm>
 #include <tuple>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/tril_indices_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/trunc_grad_kernel.cu b/paddle/phi/kernels/gpu/trunc_grad_kernel.cu
index 92d95e7259bf0c5cb1cf7c180eeba1d7b7ea8842..927978339181d1a86fe17bf047c02be6af800065 100644
--- a/paddle/phi/kernels/gpu/trunc_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/trunc_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/trunc_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/trunc_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/trunc_kernel.cu b/paddle/phi/kernels/gpu/trunc_kernel.cu
index cc44602b657aabfe7a6f55ed1d0b2c06cb56fa9e..1f374714a6ff1b9acc630563e59ccc8e494643a8 100644
--- a/paddle/phi/kernels/gpu/trunc_kernel.cu
+++ b/paddle/phi/kernels/gpu/trunc_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/trunc_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/trunc_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/truncated_gaussian_random_kernel.cu b/paddle/phi/kernels/gpu/truncated_gaussian_random_kernel.cu
index 7c59b39890f4676822789028db890e8ff8d3e46b..698dcc20ad3feb92990bc19a91848a96dbb14fcc 100644
--- a/paddle/phi/kernels/gpu/truncated_gaussian_random_kernel.cu
+++ b/paddle/phi/kernels/gpu/truncated_gaussian_random_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/truncated_gaussian_random_kernel.h"
+
 #include <thrust/device_vector.h>
 #include <thrust/host_vector.h>
 #include <thrust/random.h>
@@ -22,7 +24,6 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/truncated_gaussian_random_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/unbind_kernel.cu b/paddle/phi/kernels/gpu/unbind_kernel.cu
index 1efc3a1094da253c27fec5108b536837d868425e..8a7aa8f6033ab9b86f87e792bc37f912562578a7 100644
--- a/paddle/phi/kernels/gpu/unbind_kernel.cu
+++ b/paddle/phi/kernels/gpu/unbind_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unbind_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unbind_kernel_impl.h"
-#include "paddle/phi/kernels/unbind_kernel.h"
 
 PD_REGISTER_KERNEL(unbind,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/unfold_grad_kernel.cu b/paddle/phi/kernels/gpu/unfold_grad_kernel.cu
index 3740f59603bef4b8a245101f1ecf636d4d8423c2..9cd874ec6a6f79fa00ff147e7d4aba3085fc8f83 100644
--- a/paddle/phi/kernels/gpu/unfold_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/unfold_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unfold_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unfold_grad_kernel_impl.h"
-#include "paddle/phi/kernels/unfold_grad_kernel.h"
 
 PD_REGISTER_KERNEL(
     unfold_grad, GPU, ALL_LAYOUT, phi::UnfoldGradKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/unfold_kernel.cu b/paddle/phi/kernels/gpu/unfold_kernel.cu
index 4f72a6f794e5f3b78e503d3cc27de2de6802f701..db0d259791b60cb162a1dd39b0e44e71fef1b512 100644
--- a/paddle/phi/kernels/gpu/unfold_kernel.cu
+++ b/paddle/phi/kernels/gpu/unfold_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unfold_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unfold_kernel_impl.h"
-#include "paddle/phi/kernels/unfold_kernel.h"
 
 PD_REGISTER_KERNEL(unfold, GPU, ALL_LAYOUT, phi::UnfoldKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/uniform_random_kernel.cu b/paddle/phi/kernels/gpu/uniform_random_kernel.cu
index b149110bddce69ab7c6059bb1ede64f8f267ae16..ee486976a450c2f7c756130f78ee686c09dcd218 100644
--- a/paddle/phi/kernels/gpu/uniform_random_kernel.cu
+++ b/paddle/phi/kernels/gpu/uniform_random_kernel.cu
@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/uniform_random_kernel.h"
+
 #include <thrust/random.h>
 
 #include "gflags/gflags.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/distribution_helper.h"
 #include "paddle/phi/kernels/funcs/index_impl.cu.h"
-#include "paddle/phi/kernels/uniform_random_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/unique_kernel.cu b/paddle/phi/kernels/gpu/unique_kernel.cu
index fe9304d0e150e55b684bcc4fae9bcd054ceaf483..2f24a44c232035df5827bdeebab1e43d1acdc507 100644
--- a/paddle/phi/kernels/gpu/unique_kernel.cu
+++ b/paddle/phi/kernels/gpu/unique_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unique_kernel.h"
+
 #include <thrust/adjacent_difference.h>
 #include <thrust/device_vector.h>
 #include <thrust/execution_policy.h>
@@ -28,7 +30,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/unique_functor.h"
-#include "paddle/phi/kernels/unique_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/unsqueeze_grad_kernel.cu b/paddle/phi/kernels/gpu/unsqueeze_grad_kernel.cu
index 6bd3f30779426755a8d487cb1b96c1c466e0bcf9..6c3a2066f0f2d18f35e8c2e58b9a26ae3b5624d3 100644
--- a/paddle/phi/kernels/gpu/unsqueeze_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/unsqueeze_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unsqueeze_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unsqueeze_grad_kernel_impl.h"
-#include "paddle/phi/kernels/unsqueeze_grad_kernel.h"
 
 PD_REGISTER_KERNEL(unsqueeze_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/unsqueeze_kernel.cu b/paddle/phi/kernels/gpu/unsqueeze_kernel.cu
index 78f1fafda9bccce55d0d065007f6c9a069229351..86b44622546375321086216ff5a8b8c5b3429f55 100644
--- a/paddle/phi/kernels/gpu/unsqueeze_kernel.cu
+++ b/paddle/phi/kernels/gpu/unsqueeze_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unsqueeze_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unsqueeze_kernel_impl.h"
-#include "paddle/phi/kernels/unsqueeze_kernel.h"
 
 PD_REGISTER_KERNEL(unsqueeze,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/unstack_grad_kernel.cu b/paddle/phi/kernels/gpu/unstack_grad_kernel.cu
index 6245e8258f8bb25c730e5ea397d6deb911d1456c..b7c349de0df32b0a0687264c7504e600306068fd 100644
--- a/paddle/phi/kernels/gpu/unstack_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/unstack_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unstack_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unstack_grad_kernel_impl.h"
-#include "paddle/phi/kernels/unstack_grad_kernel.h"
 
 PD_REGISTER_KERNEL(unstack_grad,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/unstack_kernel.cu b/paddle/phi/kernels/gpu/unstack_kernel.cu
index dde6d926b7ca0c3518a8ea4ac25df04e570d69a2..f147f4c0f0edfd4af38070f3c221cdd70aef7c9c 100644
--- a/paddle/phi/kernels/gpu/unstack_kernel.cu
+++ b/paddle/phi/kernels/gpu/unstack_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/unstack_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/unstack_kernel_impl.h"
-#include "paddle/phi/kernels/unstack_kernel.h"
 
 PD_REGISTER_KERNEL(unstack,
                    GPU,
diff --git a/paddle/phi/kernels/gpu/warpctc_grad_kernel.cu b/paddle/phi/kernels/gpu/warpctc_grad_kernel.cu
index 97a75bf684911266453eec3e32cdb1a6df8d44d8..612b03555c6f16956bc350d19b99eca5f79bde32 100644
--- a/paddle/phi/kernels/gpu/warpctc_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/warpctc_grad_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/warpctc_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/warpctc_grad_kernel_impl.h"
-#include "paddle/phi/kernels/warpctc_grad_kernel.h"
 
 PD_REGISTER_KERNEL(
     warpctc_grad, GPU, ALL_LAYOUT, phi::WarpctcGradKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/warpctc_kernel.cu b/paddle/phi/kernels/gpu/warpctc_kernel.cu
index 8d93f24b659141248ca70d4b21e2c88a0d970f33..3379322f3dfd844c470046dc7c04f3d0624fd01f 100644
--- a/paddle/phi/kernels/gpu/warpctc_kernel.cu
+++ b/paddle/phi/kernels/gpu/warpctc_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/warpctc_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/warpctc_kernel_impl.h"
-#include "paddle/phi/kernels/warpctc_kernel.h"
 
 PD_REGISTER_KERNEL(
     warpctc, GPU, ALL_LAYOUT, phi::WarpctcKernel, float, double) {}
diff --git a/paddle/phi/kernels/gpu/where_grad_kernel.cu b/paddle/phi/kernels/gpu/where_grad_kernel.cu
index 2be698c3455d5a8c0ffc10fc11b5ce043a739d90..709dddcb82c7e76940d850750916758a31ea4f44 100644
--- a/paddle/phi/kernels/gpu/where_grad_kernel.cu
+++ b/paddle/phi/kernels/gpu/where_grad_kernel.cu
@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/where_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/where_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/where_kernel.cu b/paddle/phi/kernels/gpu/where_kernel.cu
index c623b6ec8b7218baf02aa7f8bdcf3c523256ec73..441be02b99efa266c55590a79b21263f05ab46d8 100644
--- a/paddle/phi/kernels/gpu/where_kernel.cu
+++ b/paddle/phi/kernels/gpu/where_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/where_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/broadcast_function.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
-#include "paddle/phi/kernels/where_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpu/yolo_box_kernel.cu b/paddle/phi/kernels/gpu/yolo_box_kernel.cu
index 2719dcd9e54094f2b6af0ad18eabb445081d60a6..446a21e6a8070ed1a0f4ccf3a0c66f23e10d4138 100644
--- a/paddle/phi/kernels/gpu/yolo_box_kernel.cu
+++ b/paddle/phi/kernels/gpu/yolo_box_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/yolo_box_kernel.h"
+
 #include "paddle/fluid/memory/malloc.h"
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -19,7 +21,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/yolo_box_util.h"
-#include "paddle/phi/kernels/yolo_box_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu b/paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu
index 13d85ed1d36bd3f63de7148c4f4c76b71346fe9a..ef70907b59a6173d876050ebdc72da76aa3c40d1 100644
--- a/paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_grad_grad_kernel.h"
+
 #include "paddle/fluid/framework/eigen.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_grad_grad_kernel.h"
 #ifdef PADDLE_WITH_HIP
 #include "paddle/fluid/operators/conv_miopen_helper.h"
 #else
diff --git a/paddle/phi/kernels/gpudnn/conv_grad_kernel.cu b/paddle/phi/kernels/gpudnn/conv_grad_kernel.cu
index 80100ba8ff44d8e179a4ecbb97b5dc567de025e2..4e9c37879c0029f2d93b76513fcf70e45031159e 100644
--- a/paddle/phi/kernels/gpudnn/conv_grad_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/conv_grad_kernel.cu
@@ -12,11 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_grad_kernel.h"
+
 #include "paddle/fluid/framework/eigen.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_grad_kernel.h"
 #ifdef PADDLE_WITH_HIP
 #include "paddle/fluid/operators/conv_miopen_helper.h"
 #else
diff --git a/paddle/phi/kernels/gpudnn/conv_kernel.cu b/paddle/phi/kernels/gpudnn/conv_kernel.cu
index c746c4db9dce8b18123accf7d40bc357073b5c8e..bd95a32bc724f2849326d43a72e8fd64175c3d22 100644
--- a/paddle/phi/kernels/gpudnn/conv_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/conv_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/conv_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_kernel.h"
 
 #ifdef PADDLE_WITH_HIP
 #include "paddle/fluid/operators/conv_miopen_helper.h"
diff --git a/paddle/phi/kernels/gpudnn/conv_transpose_grad_kernel.cu b/paddle/phi/kernels/gpudnn/conv_transpose_grad_kernel.cu
index 6d5a0dd5e0b7e8d908305ab55841cc588a504ff4..0ce16f66becfaf92552e3c42be1b940f5506639b 100644
--- a/paddle/phi/kernels/gpudnn/conv_transpose_grad_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/conv_transpose_grad_kernel.cu
@@ -12,13 +12,14 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/conv_transpose_grad_kernel.h"
+
 #include <algorithm>
 
 #include "paddle/phi/backends/dynload/cudnn.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/ddim.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_transpose_grad_kernel.h"
 #include "paddle/phi/kernels/cpu/conv_util.h"
 #include "paddle/phi/kernels/funcs/batch_norm_utils.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
diff --git a/paddle/phi/kernels/gpudnn/conv_transpose_kernel.cu b/paddle/phi/kernels/gpudnn/conv_transpose_kernel.cu
index 67a2f381d76f4db76438b6cb8df9e185aa068bda..58ead4c3287f884bb4a733b4c0c7bbd95f9f41e3 100644
--- a/paddle/phi/kernels/gpudnn/conv_transpose_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/conv_transpose_kernel.cu
@@ -12,13 +12,14 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/conv_transpose_kernel.h"
+
 #include <algorithm>
 
 #include "paddle/phi/backends/dynload/cudnn.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/ddim.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/conv_transpose_kernel.h"
 #include "paddle/phi/kernels/cpu/conv_util.h"
 #include "paddle/phi/kernels/funcs/padding.h"
 #include "paddle/phi/kernels/funcs/slice.h"
diff --git a/paddle/phi/kernels/gpudnn/pool_grad_kernel.cu b/paddle/phi/kernels/gpudnn/pool_grad_kernel.cu
index 919a2a2193a4ff17356edcdc102502cd8120050b..4a7d417f2b6b28f23482d893e1122fec57be7b91 100644
--- a/paddle/phi/kernels/gpudnn/pool_grad_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/pool_grad_kernel.cu
@@ -12,13 +12,14 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/pool_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_dnn.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/pooling.h"
 #include "paddle/phi/kernels/gpudnn/pool_gpudnn.h"
-#include "paddle/phi/kernels/pool_grad_kernel.h"
 #include "paddle/phi/kernels/pool_kernel.h"
 
 #ifdef PADDLE_WITH_HIP
diff --git a/paddle/phi/kernels/gpudnn/pool_kernel.cu b/paddle/phi/kernels/gpudnn/pool_kernel.cu
index 53e3d7c9426c0d3ce200c29cd926ce2690dc6e01..9981ac5b53e9f3835edae3c9b1ea0c117ed04400 100644
--- a/paddle/phi/kernels/gpudnn/pool_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/pool_kernel.cu
@@ -12,12 +12,13 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/pool_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_dnn.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/pooling.h"
 #include "paddle/phi/kernels/gpudnn/pool_gpudnn.h"
-#include "paddle/phi/kernels/pool_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu b/paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu
index 343cba311ae467631b58fd45c7e4d06c419b99bc..45ab645d3736734fb9ec4c6a7b949274c1f0a91e 100644
--- a/paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/softmax_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpudnn/softmax_gpudnn.h"
-#include "paddle/phi/kernels/softmax_grad_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/gpudnn/softmax_kernel.cu b/paddle/phi/kernels/gpudnn/softmax_kernel.cu
index b71f39722c98dacc3a7e16a6ca19a7391446c995..37175c427ffe142c31b41c8356d160d203fd6d73 100644
--- a/paddle/phi/kernels/gpudnn/softmax_kernel.cu
+++ b/paddle/phi/kernels/gpudnn/softmax_kernel.cu
@@ -12,10 +12,11 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/softmax_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/gpudnn/softmax_gpudnn.h"
-#include "paddle/phi/kernels/softmax_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/selected_rows/gpu/adam_kernel.cu b/paddle/phi/kernels/selected_rows/gpu/adam_kernel.cu
index ec9fed6e3d9eb5f8eb2a76bd5a29ae8bbb5ebcc4..842e05fe58e683bab89f145a17f6efdd43aca7a0 100644
--- a/paddle/phi/kernels/selected_rows/gpu/adam_kernel.cu
+++ b/paddle/phi/kernels/selected_rows/gpu/adam_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/selected_rows/adam_kernel.h"
+
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/operators/math/selected_rows_functor.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -21,7 +23,6 @@
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/adam_functors.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
-#include "paddle/phi/kernels/selected_rows/adam_kernel.h"
 
 namespace phi {
 namespace sr {
diff --git a/paddle/phi/kernels/selected_rows/gpu/adamw_kernel.cu b/paddle/phi/kernels/selected_rows/gpu/adamw_kernel.cu
index 35a349a277d7440d82257bd5ac10d599311bcd8c..6e0123d2fcaf96f38e0db29179d0a8ced7f7d71b 100644
--- a/paddle/phi/kernels/selected_rows/gpu/adamw_kernel.cu
+++ b/paddle/phi/kernels/selected_rows/gpu/adamw_kernel.cu
@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/selected_rows/adamw_kernel.h"
+
 #include <math.h>  // for sqrt in CPU and CUDA
 
 #include <vector>
@@ -25,7 +27,6 @@
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/adam_functors.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
-#include "paddle/phi/kernels/selected_rows/adamw_kernel.h"
 
 namespace phi {
 namespace sr {
diff --git a/paddle/phi/kernels/selected_rows/gpu/clip_kernel.cu b/paddle/phi/kernels/selected_rows/gpu/clip_kernel.cu
index b9f4febb3b48dee7b5c0f1208831a814ded43fd3..a8d659559e19e5c04c0145fbe918065958e6bb64 100644
--- a/paddle/phi/kernels/selected_rows/gpu/clip_kernel.cu
+++ b/paddle/phi/kernels/selected_rows/gpu/clip_kernel.cu
@@ -12,10 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include "paddle/phi/kernels/selected_rows/clip_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/float16.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/selected_rows/clip_kernel.h"
 #include "paddle/phi/kernels/selected_rows/impl/clip_kernel_impl.h"
 
 PD_REGISTER_KERNEL(clip_sr,
diff --git a/paddle/phi/kernels/sparse/gpu/coalesced_kernel.cu b/paddle/phi/kernels/sparse/gpu/coalesced_kernel.cu
index 7d9e566916addeabb3729c0587c82d02718cc315..405384009df89da78f97c80d7146f9d6fc1f5dfa 100644
--- a/paddle/phi/kernels/sparse/gpu/coalesced_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/coalesced_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/coalesced_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
@@ -20,7 +22,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/funcs/sparse/flatten_indices.cu.h"
 #include "paddle/phi/kernels/funcs/sparse/scatter.cu.h"
 #include "paddle/phi/kernels/funcs/sparse/utils.cu.h"
-#include "paddle/phi/kernels/sparse/coalesced_kernel.h"
 
 namespace phi {
 namespace sparse {
diff --git a/paddle/phi/kernels/sparse/gpu/convolution_grad_kernel.cu b/paddle/phi/kernels/sparse/gpu/convolution_grad_kernel.cu
index d83d064418eeca4cf6d2f786835225ee5332d41a..805c417b8dbc1fc70cfb3965212a65edeb2b5c32 100644
--- a/paddle/phi/kernels/sparse/gpu/convolution_grad_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/convolution_grad_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/convolution_grad_kernel.h"
+
 #include "glog/logging.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
@@ -23,7 +25,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/funcs/blas/blas.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
-#include "paddle/phi/kernels/sparse/convolution_grad_kernel.h"
 #include "paddle/phi/kernels/sparse/gpu/convolution.cu.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/sparse/gpu/convolution_kernel.cu b/paddle/phi/kernels/sparse/gpu/convolution_kernel.cu
index c3b6c8c6abcc83e4285056c4c75f0a4cc0b3fd51..fe66fb5cff9de40a4b89748a6a5ddca380371123 100644
--- a/paddle/phi/kernels/sparse/gpu/convolution_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/convolution_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/convolution_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_meta.h"
@@ -19,7 +21,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/funcs/blas/blas.h"
 #include "paddle/phi/kernels/funcs/scatter.cu.h"
 #include "paddle/phi/kernels/funcs/sparse/scatter.cu.h"
-#include "paddle/phi/kernels/sparse/convolution_kernel.h"
 #include "paddle/phi/kernels/sparse/gpu/convolution.cu.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/sparse/gpu/full_kernel.cu b/paddle/phi/kernels/sparse/gpu/full_kernel.cu
index 6c1e0160bca8e9ff4448fcc2735d97b72aae6ad0..500217d6edc502aa53b6b6f71700185b06390eea 100644
--- a/paddle/phi/kernels/sparse/gpu/full_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/full_kernel.cu
@@ -12,12 +12,13 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/full_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/dense_tensor.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/funcs/elementwise_base.h"
-#include "paddle/phi/kernels/sparse/full_kernel.h"
 
 namespace phi {
 
diff --git a/paddle/phi/kernels/sparse/gpu/matmul_grad_kernel.cu b/paddle/phi/kernels/sparse/gpu/matmul_grad_kernel.cu
index 30344c307541e6f5eb34a85323b796d00e863f1e..6db1bfac6bc41fbd0ab07fa3d9212c1bc5a221c2 100644
--- a/paddle/phi/kernels/sparse/gpu/matmul_grad_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/matmul_grad_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/matmul_grad_kernel.h"
+
 #include <vector>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -19,7 +21,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/empty_kernel.h"
 #include "paddle/phi/kernels/funcs/sparse/sparse_blas.h"
-#include "paddle/phi/kernels/sparse/matmul_grad_kernel.h"
 #include "paddle/phi/kernels/transpose_kernel.h"
 
 namespace phi {
diff --git a/paddle/phi/kernels/sparse/gpu/matmul_kernel.cu b/paddle/phi/kernels/sparse/gpu/matmul_kernel.cu
index f8590737eda074705da126619bbd8e9e870b2890..e1192687ac4a3888de1b5ded3a634e0c9ddd5db6 100644
--- a/paddle/phi/kernels/sparse/gpu/matmul_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/matmul_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/matmul_kernel.h"
+
 #include <vector>
 
 #include "paddle/phi/backends/gpu/gpu_context.h"
@@ -24,7 +26,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/empty_kernel.h"
 #include "paddle/phi/kernels/funcs/sparse/sparse_blas.h"
-#include "paddle/phi/kernels/sparse/matmul_kernel.h"
 
 namespace phi {
 namespace sparse {
diff --git a/paddle/phi/kernels/sparse/gpu/sparse_mask_kernel.cu b/paddle/phi/kernels/sparse/gpu/sparse_mask_kernel.cu
index cbbdc122f616f7d2100f4b313cc29506b35d2df0..2153d9dfe6853b3971f4697ce27c60f59aeb2d11 100644
--- a/paddle/phi/kernels/sparse/gpu/sparse_mask_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/sparse_mask_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/sparse_mask_kernel.h"
+
 #include <thrust/binary_search.h>
 
 #include "paddle/phi/backends/gpu/gpu_info.h"
@@ -24,7 +26,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/empty_kernel.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/sparse/flatten_indices.cu.h"
-#include "paddle/phi/kernels/sparse/sparse_mask_kernel.h"
 
 namespace phi {
 namespace sparse {
diff --git a/paddle/phi/kernels/sparse/gpu/sparse_pool_grad_kernel.cu b/paddle/phi/kernels/sparse/gpu/sparse_pool_grad_kernel.cu
index 694fe667c879e3ec08724bc777996cb967da5bbd..669ecb017dc02e3b1352f7ac250981adb4c2b737 100644
--- a/paddle/phi/kernels/sparse/gpu/sparse_pool_grad_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/sparse_pool_grad_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/sparse_pool_grad_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
@@ -22,7 +24,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/pooling.h"
 #include "paddle/phi/kernels/funcs/sparse/convolution.h"
-#include "paddle/phi/kernels/sparse/sparse_pool_grad_kernel.h"
 
 namespace phi {
 namespace sparse {
diff --git a/paddle/phi/kernels/sparse/gpu/sparse_pool_kernel.cu b/paddle/phi/kernels/sparse/gpu/sparse_pool_kernel.cu
index 534afbd0f142143a196cc96c0cfdf6ae788685c1..bc6723d26b7a6067ab1ef2bfa4e5e1d3a61a862f 100644
--- a/paddle/phi/kernels/sparse/gpu/sparse_pool_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/sparse_pool_kernel.cu
@@ -12,13 +12,14 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/sparse_pool_kernel.h"
+
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_meta.h"
 #include "paddle/phi/core/visit_type.h"
 #include "paddle/phi/kernels/funcs/pooling.h"
 #include "paddle/phi/kernels/funcs/sparse/convolution.h"
 #include "paddle/phi/kernels/sparse/gpu/convolution.cu.h"
-#include "paddle/phi/kernels/sparse/sparse_pool_kernel.h"
 
 namespace phi {
 namespace sparse {
diff --git a/paddle/phi/kernels/sparse/gpu/sparse_utils_kernel.cu b/paddle/phi/kernels/sparse/gpu/sparse_utils_kernel.cu
index 1ed4ebd23db87ebf353844e4b8daf029de41dba2..52b905e1bdf5a8d2bd416374fa8f056fa5e58954 100644
--- a/paddle/phi/kernels/sparse/gpu/sparse_utils_kernel.cu
+++ b/paddle/phi/kernels/sparse/gpu/sparse_utils_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/sparse/sparse_utils_kernel.h"
+
 #include <thrust/execution_policy.h>
 #include <thrust/remove.h>
 
@@ -23,7 +25,6 @@ limitations under the License. */
 #include "paddle/phi/core/visit_type.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/sparse/common_shape.h"
-#include "paddle/phi/kernels/sparse/sparse_utils_kernel.h"
 
 namespace phi {
 namespace sparse {
diff --git a/paddle/phi/kernels/strings/gpu/strings_copy_kernel.cu b/paddle/phi/kernels/strings/gpu/strings_copy_kernel.cu
index 7a2d61f29f7cb9e632119b0f5f0a8abaf08c2c85..c49b41e0d3f29b69beb4b93c2d1c2095edc1f37b 100644
--- a/paddle/phi/kernels/strings/gpu/strings_copy_kernel.cu
+++ b/paddle/phi/kernels/strings/gpu/strings_copy_kernel.cu
@@ -12,6 +12,8 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/strings/strings_copy_kernel.h"
+
 #include "paddle/phi/backends/all_context.h"
 #include "paddle/phi/backends/gpu/gpu_helper.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
@@ -20,7 +22,6 @@ limitations under the License. */
 #include "paddle/phi/kernels/copy_kernel.h"
 #include "paddle/phi/kernels/empty_kernel.h"
 #include "paddle/phi/kernels/strings/gpu/copy_utils.h"
-#include "paddle/phi/kernels/strings/strings_copy_kernel.h"
 
 using pstring = ::phi::dtype::pstring;
 
diff --git a/paddle/phi/kernels/strings/gpu/strings_lower_upper_kernel.cu b/paddle/phi/kernels/strings/gpu/strings_lower_upper_kernel.cu
index 05d868f4db831c7dd879e6a125266da764b3df8f..2237edae7eb776071c19fafbc5dd6adf36914528 100644
--- a/paddle/phi/kernels/strings/gpu/strings_lower_upper_kernel.cu
+++ b/paddle/phi/kernels/strings/gpu/strings_lower_upper_kernel.cu
@@ -9,11 +9,12 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include "paddle/phi/kernels/strings/strings_lower_upper_kernel.h"
+
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/common/pstring.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/strings/strings_lower_upper_kernel.h"
 #include "paddle/phi/kernels/strings/unicode.h"
 
 using pstring = ::phi::dtype::pstring;