diff --git a/paddle/pten/api/lib/kernel_dispatch.h b/paddle/pten/api/lib/kernel_dispatch.h
index de753669af667978f801922344ff4ab0645ab119..17ef22cb6dea9c629fc3845eb9f770cc85272539 100644
--- a/paddle/pten/api/lib/kernel_dispatch.h
+++ b/paddle/pten/api/lib/kernel_dispatch.h
@@ -46,8 +46,6 @@ enum class KernelType {
 
 // TODO(chenweihang): support DataLayout and DataType selected
 struct KernelKeySet {
-  KernelType kernel_type{KernelType::DENSE_TENSOR_KENREL};
-
   BackendSet backend_set{Backend::UNDEFINED};
   DataLayout layout{DataLayout::UNDEFINED};
   DataType dtype{DataType::UNDEFINED};
@@ -97,9 +95,6 @@ struct KernelKeyParser : ArgsIterator<KernelKeyParser> {
   void operator()(const Tensor& x) {
     key_set.backend_set = key_set.backend_set | detail::GetTensorBackendSet(x);
     // TODO(chenweihang): selecte multi layout and dtype
-    if (pten::SelectedRows::classof(x.impl().get())) {
-      key_set.kernel_type = KernelType::SELECTED_ROWS_KENREL;
-    }
     key_set.layout = x.layout();
     key_set.dtype = x.type();
     dtype_set = dtype_set | DataTypeSet(x.dtype());
@@ -124,6 +119,24 @@ struct KernelKeyParser : ArgsIterator<KernelKeyParser> {
   }
 };
 
+struct KernelTypeParser : ArgsIterator<KernelTypeParser> {
+  KernelType kernel_type{KernelType::DENSE_TENSOR_KENREL};
+
+  // TODO(chenweihang): deal with multiple diff input Tensors
+  // TODO(chenweihang): add global device guard method to set backend
+  void operator()(const Tensor& x) {
+    if (pten::SelectedRows::classof(x.impl().get())) {
+      kernel_type = KernelType::SELECTED_ROWS_KENREL;
+    }
+  }
+
+  // skip other type args, these args don't used in kernel selection
+  template <typename T>
+  void operator()(const T& x) {
+    // do nothing
+  }
+};
+
 }  // namespace detail
 
 template <typename... Args>
@@ -131,6 +144,11 @@ KernelKeySet ParseKernelKeyByInputArgs(const Args&... args) {
   return detail::KernelKeyParser().apply(args...).key_set;
 }
 
+template <typename... Args>
+KernelType ParseKernelTypeByInputArgs(const Args&... args) {
+  return detail::KernelTypeParser().apply(args...).kernel_type;
+}
+
 DataType ParseDataType(DataType dtype);
 DataType ParseDataType(const Tensor& tensor);
 DataType ParseDataType(const std::vector<Tensor>& tensors);
diff --git a/python/paddle/utils/code_gen/api_base.py b/python/paddle/utils/code_gen/api_base.py
index 73c3ba4e4b4fe9d56ac6e6c7638777fc5df89164..7515981490728774de215c6c8698f96884cd946a 100644
--- a/python/paddle/utils/code_gen/api_base.py
+++ b/python/paddle/utils/code_gen/api_base.py
@@ -345,7 +345,7 @@ PADDLE_API {self.outputs['return_type']} {self.api}({self.args_str['args_declare
         if len(input_names) > 0:
             if self.support_selected_rows_kernel:
                 kernel_select_code = kernel_select_code + f"""
-  KernelType kernel_type;
+  KernelType kernel_type = ParseKernelTypeByInputArgs({", ".join(input_names)});
 """
 
             kernel_select_code = kernel_select_code + f"""
@@ -354,7 +354,6 @@ PADDLE_API {self.outputs['return_type']} {self.api}({self.args_str['args_declare
         || kernel_data_type == DataType::UNDEFINED ) {{
     auto kernel_key_set = ParseKernelKeyByInputArgs({kernel_select_args});
     auto kernel_key = kernel_key_set.GetHigestPriorityKernelKey();
-    {'kernel_type = kernel_key_set.kernel_type;' if self.support_selected_rows_kernel else ''}
     if (kernel_backend == Backend::UNDEFINED) {{
       kernel_backend = kernel_key.backend();
     }}