diff --git a/paddle/pten/api/lib/CMakeLists.txt b/paddle/pten/api/lib/CMakeLists.txt
index 2cf737eb8b17f0d60eaae40c465e612698a64e6c..969ac517513cc584359881179224bd8a32e09d9e 100644
--- a/paddle/pten/api/lib/CMakeLists.txt
+++ b/paddle/pten/api/lib/CMakeLists.txt
@@ -3,19 +3,13 @@ add_subdirectory(utils)
 cc_library(ext_compat_utils SRCS ext_compat_utils.cc DEPS place)
 
 if (WITH_GPU)
-  nv_library(pten_tensor SRCS tensor.cc DEPS tensor_base dense_tensor pten_api_utils ext_compat_utils pten_enforce manual_api pten_function_api)
+  nv_library(pten_tensor_raw SRCS tensor.cc DEPS tensor_base dense_tensor pten_api_utils ext_compat_utils pten_enforce manual_api)
 elseif (WITH_ROCM)
-  hip_library(pten_tensor SRCS tensor.cc DEPS tensor_base dense_tensor pten_api_utils ext_compat_utils pten_enforce manual_api pten_function_api)
+  hip_library(pten_tensor_raw SRCS tensor.cc DEPS tensor_base dense_tensor pten_api_utils ext_compat_utils pten_enforce manual_api)
 else()
-  cc_library(pten_tensor SRCS tensor.cc DEPS tensor_base dense_tensor pten_api_utils ext_compat_utils pten_enforce manual_api pten_function_api)
+  cc_library(pten_tensor_raw SRCS tensor.cc DEPS tensor_base dense_tensor pten_api_utils ext_compat_utils pten_enforce manual_api)
 endif()
 
-cc_library(kernel_dispatch SRCS kernel_dispatch.cc DEPS pten_tensor pten_context kernel_factory)
-
-cc_library(op_meta_info SRCS op_meta_info.cc DEPS pten_tensor)
-cc_library(op_kernel_info SRCS op_kernel_info.cc DEPS pten_tensor)
-
-
 set(api_gen_base ${CMAKE_SOURCE_DIR}/python/paddle/utils/code_gen/api_base.py)
 
 # forward api file
@@ -81,8 +75,15 @@ add_custom_command(
   DEPENDS ${api_yaml_file} ${wrapped_infermeta_gen_file} ${api_gen_base}
   VERBATIM)
 
-cc_library(pten_data_transform SRCS data_transform.cc DEPS pten_tensor transfer_layout_kernel cast_kernel data_device_transform)
-cc_library(manual_api SRCS manual_api.cc DEPS pten_tensor pten kernel_dispatch)
+cc_library(kernel_dispatch SRCS kernel_dispatch.cc DEPS pten_tensor_raw pten_context kernel_factory)
+cc_library(pten_data_transform SRCS data_transform.cc DEPS pten_tensor_raw transfer_layout_kernel cast_kernel data_device_transform)
+cc_library(manual_api SRCS manual_api.cc DEPS pten_tensor_raw pten kernel_dispatch pten_data_transform)
+
+cc_library(pten_tensor SRCS tensor_method.cc DEPS pten_tensor_raw pten_function_api)
+
+cc_library(op_meta_info SRCS op_meta_info.cc DEPS pten_tensor)
+cc_library(op_kernel_info SRCS op_kernel_info.cc DEPS pten_tensor_raw)
+
 cc_library(sparse_api SRCS sparse_api.cc DEPS pten_tensor pten kernel_dispatch pten_data_transform)
 cc_library(pten_function_api SRCS ${api_source_file} DEPS pten_tensor pten kernel_dispatch pten_data_transform)
 cc_library(pten_bw_function_api SRCS ${bw_api_source_file} DEPS pten_tensor pten kernel_dispatch backward_infermeta pten_data_transform pten_function_api)
diff --git a/paddle/pten/api/lib/manual_api.cc b/paddle/pten/api/lib/manual_api.cc
index 667bd177ee1f6232f44299502e348e23579cb49c..8ee6115e982f54fd3261afd7e378ab9d8df5e0c6 100644
--- a/paddle/pten/api/lib/manual_api.cc
+++ b/paddle/pten/api/lib/manual_api.cc
@@ -28,6 +28,7 @@ limitations under the License. */
 #include "paddle/pten/infermeta/unary.h"
 
 PT_DECLARE_KERNEL(copy, CPU, ALL_LAYOUT);
+PT_DECLARE_KERNEL(split, CPU, ALL_LAYOUT);
 
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
 PT_DECLARE_KERNEL(copy, GPU, ALL_LAYOUT);
diff --git a/paddle/pten/api/lib/tensor.cc b/paddle/pten/api/lib/tensor.cc
index aae11294b0be0899480504fd734d44e584abbe5b..cc68b19bf887196770e1729a12a6885751d1f1ff 100644
--- a/paddle/pten/api/lib/tensor.cc
+++ b/paddle/pten/api/lib/tensor.cc
@@ -58,9 +58,6 @@ limitations under the License. */
 namespace paddle {
 namespace experimental {
 
-// declare cast api
-Tensor cast(const Tensor &x, DataType out_dtype);
-
 /////// Tensor Methods ////////
 
 /* Part 1: Construction and destruction methods */
@@ -363,9 +360,6 @@ void Tensor::copy_(const Tensor &src, bool blocking) {
       src.copy_to(pten::TransToPtenBackend(src.inner_place()), blocking);
   set_impl(copy_tensor.impl());
 }
-Tensor Tensor::cast(DataType target_type) const {
-  return experimental::cast(*this, target_type);
-}
 
 /* Part 6: Status utils methods */
 
diff --git a/paddle/pten/api/lib/tensor_method.cc b/paddle/pten/api/lib/tensor_method.cc
new file mode 100644
index 0000000000000000000000000000000000000000..35d50b9e668b45624d826da75fa28bbdd84ddf60
--- /dev/null
+++ b/paddle/pten/api/lib/tensor_method.cc
@@ -0,0 +1,28 @@
+/* Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+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/pten/api/include/tensor.h"
+
+namespace paddle {
+namespace experimental {
+
+// declare cast api
+Tensor cast(const Tensor &x, DataType out_dtype);
+
+Tensor Tensor::cast(DataType target_type) const {
+  return experimental::cast(*this, target_type);
+}
+
+}  // namespace experimental
+}  // namespace paddle