Fix some problem of kernel fallback in C++ API (#44681)

* support auto fallback to  cpu kernel for cusom device

* fix some problem of kernel fallback

paddle/phi/api/lib/api_custom_impl.cc

@@ -73,8 +73,9 @@ std::tuple<Tensor, Tensor, Tensor, Tensor, Tensor, Tensor> adamw_impl(
     }
   }
   std::string kernel_name = "adamw";
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       kernel_name, {kernel_backend, kernel_layout, kernel_data_type});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << kernel_name << " API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
   VLOG(6) << kernel_name << " API kernel: " << kernel;
@@ -232,8 +233,9 @@ Tensor conv2d_impl(const Tensor& input,
 
   VLOG(6) << "conv2d API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "conv2d", {kernel_backend, kernel_layout, kernel_data_type}, true);
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "conv2d API kernel: " << kernel;
 
   auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
@@ -334,8 +336,9 @@ Tensor conv3d_impl(const Tensor& input,
 
   VLOG(6) << "conv3d API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "conv3d", {kernel_backend, kernel_layout, kernel_data_type}, true);
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "conv3d API kernel: " << kernel;
 
   auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
@@ -437,8 +440,9 @@ void conv2d_grad_impl(const Tensor& input,
 
   VLOG(6) << "conv2d_grad API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "conv2d_grad", {kernel_backend, kernel_layout, kernel_data_type}, true);
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "conv2d_grad API kernel: " << kernel;
 
   auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
@@ -538,8 +542,9 @@ void conv3d_grad_impl(const Tensor& input,
 
   VLOG(6) << "conv3d_grad API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "conv3d_grad", {kernel_backend, kernel_layout, kernel_data_type}, true);
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "conv3d_grad API kernel: " << kernel;
 
   auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
@@ -624,10 +629,11 @@ Tensor embedding_impl(const Tensor& x,
   Tensor api_output;
 
   if (phi::DenseTensor::classof(weight.impl().get())) {
-    const auto& kernel =
+    auto kernel_result =
         phi::KernelFactory::Instance().SelectKernelOrThrowError(
             "embedding",
             {kernel_key.backend(), kernel_key.layout(), kernel_data_type});
+    const auto& kernel = kernel_result.kernel;
     VLOG(6) << "embedding API kernel: " << kernel;
 
     auto input_x = PrepareData(x, kernel.InputAt(0), {});
@@ -652,10 +658,11 @@ Tensor embedding_impl(const Tensor& x,
       (*kernel_fn)(*dev_ctx, *input_x, *input_weight, padding_idx, kernel_out);
     }
   } else {
-    const auto& kernel =
+    auto kernel_result =
         phi::KernelFactory::Instance().SelectKernelOrThrowError(
             "sparse_weight_embedding",
             {kernel_key.backend(), kernel_key.layout(), kernel_data_type});
+    const auto& kernel = kernel_result.kernel;
     VLOG(6) << "sparse_weight_embedding API kernel: " << kernel;
 
     auto input_x = PrepareData(x, kernel.InputAt(0), {});
@@ -693,8 +700,9 @@ std::vector<Tensor> split_impl(const Tensor& x,
   DataLayout kernel_layout = kernel_key.layout();
   DataType kernel_data_type = kernel_key.dtype();
 
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "split", {kernel_backend, kernel_layout, kernel_data_type});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "split API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
   VLOG(6) << "split API kernel: " << kernel;
@@ -774,8 +782,9 @@ std::tuple<Tensor, Tensor, Tensor> momentum_impl(
   if (grad.is_selected_rows()) {
     kernel_name = "momentum_dense_param_sparse_grad";
   }
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       kernel_name, {kernel_backend, kernel_layout, kernel_data_type});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << kernel_name << " API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
   VLOG(6) << kernel_name << " API kernel: " << kernel;
@@ -906,8 +915,9 @@ std::tuple<Tensor, Tensor, Tensor, Tensor, Tensor, Tensor> batch_norm_impl(
     }
   }
 
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "batch_norm", {kernel_backend, kernel_layout, kernel_data_type});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "batch_norm API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
   VLOG(6) << "batch_norm API kernel: " << kernel;
@@ -1004,8 +1014,9 @@ void imag_grad_impl(const Tensor& out_grad, Tensor* x_grad) {
   phi::KernelKey kernel_key{ParseBackend(out_grad),
                             out_grad.layout(),
                             phi::dtype::ToComplex(out_grad.dtype())};
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "imag_grad", kernel_key);
+  const auto& kernel = kernel_result.kernel;
 
   VLOG(6) << "imag_grad API kernel key: " << kernel_key;
   VLOG(6) << "imag_grad API kernel: " << kernel;
@@ -1042,10 +1053,11 @@ void embedding_grad_impl(const Tensor& x,
   if (phi::DenseTensor::classof(weight.impl().get())) {
     std::string kernel_name =
         sparse ? "embedding_sparse_grad" : "embedding_grad";
-    const auto& kernel =
+    auto kernel_result =
         phi::KernelFactory::Instance().SelectKernelOrThrowError(
             kernel_name,
             {kernel_key.backend(), kernel_key.layout(), kernel_data_type});
+    const auto& kernel = kernel_result.kernel;
     VLOG(6) << kernel_name << " API kernel: " << kernel;
 
     auto input_x = PrepareData(x, kernel.InputAt(0), {});
@@ -1094,10 +1106,11 @@ void embedding_grad_impl(const Tensor& x,
   } else {
     std::string kernel_name = sparse ? "sparse_weight_embedding_sparse_grad"
                                      : "sparse_weight_embedding_grad";
-    const auto& kernel =
+    auto kernel_result =
         phi::KernelFactory::Instance().SelectKernelOrThrowError(
             kernel_name,
             {kernel_key.backend(), kernel_key.layout(), kernel_data_type});
+    const auto& kernel = kernel_result.kernel;
     VLOG(6) << kernel_name << " API kernel: " << kernel;
 
     auto input_x = PrepareData(x, kernel.InputAt(0), {});
@@ -1148,8 +1161,9 @@ void real_grad_impl(const Tensor& out_grad, Tensor* x_grad) {
   phi::KernelKey kernel_key{ParseBackend(out_grad),
                             out_grad.layout(),
                             phi::dtype::ToComplex(out_grad.dtype())};
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "real_grad", kernel_key);
+  const auto& kernel = kernel_result.kernel;
 
   VLOG(6) << "real_grad API kernel key: " << kernel_key;
   VLOG(6) << "real_grad API kernel: " << kernel;

paddle/phi/api/lib/data_transform.cc

@@ -294,5 +294,30 @@ paddle::optional<std::vector<phi::DenseTensor>> PrepareData(
   return paddle::none;
 }
 
+void TransDataBackend(const phi::DenseTensor* tensor,
+                      Backend target_backend,
+                      phi::DenseTensor* out) {
+  if (tensor) {
+    *out = TransDataPlace(*tensor, phi::TransToPhiPlace(target_backend));
+  }
+}
+
+void TransDataBackend(const std::vector<phi::DenseTensor*>& tensors,
+                      Backend target_backend,
+                      std::vector<phi::DenseTensor*> outs) {
+  size_t n = tensors.size();
+  for (size_t i = 0; i < n; ++i) {
+    TransDataBackend(tensors[i], target_backend, outs[i]);
+  }
+}
+
+void TransDataBackend(const phi::SelectedRows* tensor,
+                      Backend target_backend,
+                      phi::SelectedRows* out) {
+  if (tensor) {
+    TransDataBackend(&tensor->value(), target_backend, out->mutable_value());
+  }
+}
+
 }  // namespace experimental
 }  // namespace paddle

paddle/phi/api/lib/data_transform.h

@@ -16,6 +16,7 @@ limitations under the License. */
 
 #include "paddle/phi/api/include/tensor.h"
 #include "paddle/phi/core/kernel_factory.h"
+#include "paddle/phi/core/selected_rows.h"
 
 namespace paddle {
 namespace experimental {
@@ -81,5 +82,17 @@ paddle::optional<std::vector<phi::DenseTensor>> PrepareData(
     const phi::TensorArgDef& target_args_def,
     const TransformFlag& transform_flag);
 
+void TransDataBackend(const phi::DenseTensor* tensor,
+                      Backend target_backend,
+                      phi::DenseTensor* out);
+
+void TransDataBackend(const std::vector<phi::DenseTensor*>& tensor,
+                      Backend target_backend,
+                      std::vector<phi::DenseTensor*> out);
+
+void TransDataBackend(const phi::SelectedRows* tensor,
+                      Backend target_backend,
+                      phi::SelectedRows* out);
+
 }  // namespace experimental
 }  // namespace paddle

paddle/phi/api/lib/sparse_api_custom_impl.cc

@@ -38,8 +38,9 @@ Tensor to_sparse_coo_impl(const Tensor& x, const int64_t sparse_dim) {
   auto kernel_key_set = ParseKernelKeyByInputArgs(x);
   auto kernel_key = kernel_key_set.GetHighestPriorityKernelKey();
 
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       kernel_name, kernel_key);
+  const auto& kernel = kernel_result.kernel;
 
   VLOG(6) << "add API kernel key: " << kernel_key;
   VLOG(6) << "to API kernel: " << kernel;
@@ -95,8 +96,9 @@ Tensor to_sparse_csr_impl(const Tensor& x) {
   auto kernel_key_set = ParseKernelKeyByInputArgs(x);
   auto kernel_key = kernel_key_set.GetHighestPriorityKernelKey();
 
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       kernel_name, kernel_key);
+  const auto& kernel = kernel_result.kernel;
 
   VLOG(6) << "add API kernel key: " << kernel_key;
   VLOG(6) << "to API kernel: " << kernel;
@@ -157,8 +159,9 @@ Tensor to_dense_impl(const Tensor& x) {
   auto kernel_key_set = ParseKernelKeyByInputArgs(x);
   auto kernel_key = kernel_key_set.GetHighestPriorityKernelKey();
 
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       kernel_name, kernel_key);
+  const auto& kernel = kernel_result.kernel;
 
   VLOG(6) << "add API kernel key: " << kernel_key;
   VLOG(6) << "to API kernel: " << kernel;

paddle/phi/api/yaml/generator/api_base.py

@@ -691,15 +691,21 @@ PADDLE_API {self.get_return_type(inplace_flag=True)} {api_func_name}({self.get_d
         outputs_args, kernel_output_names, output_create = self.gene_output(
             self.outputs['types'], out_tensor_type_list, code_indent,
             inplace_flag)
+        fallback_kernel_output_trans = ""
+        for kernel_out in outputs_args:
+            fallback_kernel_output_trans += (f"""
+{code_indent}    TransDataBackend({kernel_out}, kernel_backend, {kernel_out});"""
+                                             )
         cudnn_args = '' if self.kernel[
             'use_gpudnn'] == 'false' else ', ' + self.kernel['use_gpudnn']
         return f"""
 {code_indent}  VLOG(6) << "{self.api} API kernel key: [" << kernel_backend << ", " << kernel_layout << ", "<< kernel_data_type << "]";
-{code_indent}  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+{code_indent}  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
 {code_indent}      "{kernel_name}", {{kernel_backend, kernel_layout, kernel_data_type}}{cudnn_args});
+{code_indent}  const auto& kernel = kernel_result.kernel;
 {code_indent}  VLOG(6) << "{kernel_name} kernel: " << kernel;
 
-{code_indent}  auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
+{code_indent}  auto* dev_ctx = GetDeviceContextByBackend(kernel_result.has_fallback_cpu ? Backend::CPU : kernel_backend);
 {input_tensors}
 {output_create}
 {self.gene_infer_meta(kernel_output_names, code_indent)}
@@ -708,7 +714,10 @@ PADDLE_API {self.get_return_type(inplace_flag=True)} {api_func_name}({self.get_d
 {code_indent}  auto* kernel_fn = kernel.GetVariadicKernelFn<kernel_signature>();
 {code_indent}  {{
 {code_indent}    paddle::platform::RecordEvent kernel_record_event(\"{kernel_name} compute\", paddle::platform::TracerEventType::OperatorInner, 1);
-{code_indent}    (*kernel_fn)({kernel_args}, {outputs_args});
+{code_indent}    (*kernel_fn)({kernel_args}, {", ".join(outputs_args)});
+{code_indent}  }}
+{code_indent}  if (kernel_result.has_fallback_cpu) {{
+{fallback_kernel_output_trans}
 {code_indent}  }}
 
 {code_indent}  {self.gene_return_code()}"""

paddle/phi/api/yaml/generator/api_gen.py

@@ -137,13 +137,13 @@ class ForwardAPI(BaseAPI):
                     out_tensor_type_list=None,
                     code_indent='',
                     inplace_flag=False):
-        kernel_output = ""
+        kernel_output = []
         output_names = []
         output_create = ""
         return_type = self.get_return_type_with_intermediate(inplace_flag)
 
         if len(out_dtype_list) == 1:
-            kernel_output = 'kernel_out'
+            kernel_output.append('kernel_out')
             output_names.append('kernel_out')
             inplace_assign = " = " + self.inplace_map[
                 self.outputs['names'][0]] if inplace_flag and self.outputs[
@@ -186,7 +186,7 @@ class ForwardAPI(BaseAPI):
                 output_create = output_create[:-2] + '};'
 
             for i in range(len(out_dtype_list)):
-                kernel_output = kernel_output + f'kernel_out_{i}, '
+                kernel_output.append(f'kernel_out_{i}')
                 output_names.append(f'kernel_out_{i}')
                 set_out_func = 'SetKernelOutput' if out_tensor_type_list is None or out_tensor_type_list[
                     i] == 'dense' else 'SetSelectedRowsKernelOutput'
@@ -214,7 +214,6 @@ class ForwardAPI(BaseAPI):
 {code_indent}  kernel_out_{i}->ShareInplaceVersionCounterWith(*{PREFIX_TENSOR_NAME}{self.view_map[self.outputs['names'][i]]});
 {code_indent}  VLOG(3) << "Perform View between Output and Input Tensor, share allocation and inplace version.";"""
 
-            kernel_output = kernel_output[:-2]
         else:
             raise ValueError(
                 "{} : Output error: the output should not be empty.".format(

paddle/phi/api/yaml/generator/backward_api_gen.py

@@ -118,12 +118,12 @@ class BackwardAPI(BaseAPI):
                     out_tensor_type_list=None,
                     code_indent='',
                     inplace_flag=False):
-        kernel_output = ""
+        kernel_output = []
         output_names = []
         output_create = ""
 
         if len(out_dtype_list) == 1:
-            kernel_output = 'kernel_out'
+            kernel_output.append('kernel_out')
             output_names.append('kernel_out')
             inplace_assign = " = " + self.inplace_map[self.outputs['names'][
                 0]] if inplace_flag and self.inplace_map is not None and self.outputs[
@@ -144,7 +144,7 @@ class BackwardAPI(BaseAPI):
         elif len(out_dtype_list) > 1:
             output_create = ""
             for i, out_type_item in enumerate(out_dtype_list):
-                kernel_output = kernel_output + f'kernel_out_{i}, '
+                kernel_output.append(f'kernel_out_{i}')
                 output_names.append(f'kernel_out_{i}')
                 set_out_func = 'SetKernelOutput' if out_tensor_type_list is None or out_tensor_type_list[
                     i] == 'dense' else 'SetSelectedRowsKernelOutput'
@@ -168,7 +168,6 @@ class BackwardAPI(BaseAPI):
                     output_create = output_create + f"""
 {code_indent}  auto kernel_out_{i} = {set_out_func}(&{self.outputs['names'][i]});"""
 
-            kernel_output = kernel_output[:-2]
         else:
             raise ValueError(
                 "{} : Output error: the output should not be empty.".format(

paddle/phi/api/yaml/generator/sparse_api_gen.py

@@ -36,7 +36,7 @@ class SparseAPI(ForwardAPI):
                     out_tensor_type_list=None,
                     code_indent='',
                     inplace_flag=False):
-        kernel_output = ""
+        kernel_output = []
         output_names = []
         output_create = ""
         return_type = self.get_return_type_with_intermediate(inplace_flag)
@@ -47,7 +47,7 @@ class SparseAPI(ForwardAPI):
         }
 
         if len(out_dtype_list) == 1:
-            kernel_output = 'kernel_out'
+            kernel_output.append('kernel_out')
             output_names.append('kernel_out')
             inplace_assign = " = " + self.inplace_map[self.outputs['names'][
                 0]] if inplace_flag and self.inplace_map is not None and self.outputs[
@@ -73,12 +73,11 @@ class SparseAPI(ForwardAPI):
                 output_create = output_create[:-2] + '};'
 
             for i in range(len(out_dtype_list)):
-                kernel_output = kernel_output + f'kernel_out_{i}, '
+                kernel_output.append(f'kernel_out_{i}')
                 output_names.append(f'kernel_out_{i}')
                 output_create = output_create + f"""
     auto* kernel_out_{i} = SetSparseKernelOutput(&std::get<{i}>(api_output), {output_type_map[out_dtype_list[i]]});"""
 
-            kernel_output = kernel_output[:-2]
         else:
             raise ValueError(
                 "{} : Output error: the output should not be empty.".format(
@@ -147,11 +146,12 @@ class SparseAPI(ForwardAPI):
             self.gene_return_code()) == 0 else "  " + self.gene_return_code()
         return f"""
     VLOG(6) << "{self.api} api sparse kernel key: [" << kernel_backend << ", " << kernel_layout << ", "<< kernel_data_type << "]";
-    auto phi_kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+    auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
         "{kernel_name}", {{kernel_backend, kernel_layout, kernel_data_type}});
+    const auto& phi_kernel = kernel_result.kernel;
     VLOG(6) << "{self.api} api sparse kernel: " << phi_kernel;
 
-    auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
+    auto* dev_ctx = GetDeviceContextByBackend(kernel_result.has_fallback_cpu ? Backend::CPU : kernel_backend);
     auto kernel_context = phi::KernelContext(dev_ctx);
 {output_create}
 {kernel_context_code}

paddle/phi/api/yaml/generator/sparse_bw_api_gen.py

@@ -52,7 +52,7 @@ class SparseBackwardAPI(SparseAPI, BackwardAPI):
                     out_tensor_type_list=None,
                     code_indent='',
                     inplace_flag=False):
-        kernel_output = ""
+        kernel_output = []
         output_names = []
         output_create = ""
         output_type_map = {
@@ -62,7 +62,7 @@ class SparseBackwardAPI(SparseAPI, BackwardAPI):
         }
 
         if len(out_dtype_list) == 1:
-            kernel_output = 'kernel_out'
+            kernel_output.append('kernel_out')
             output_names.append('kernel_out')
             inplace_assign = " = " + self.inplace_map[self.outputs['names'][
                 0]] if inplace_flag and self.inplace_map is not None and self.outputs[
@@ -74,7 +74,7 @@ class SparseBackwardAPI(SparseAPI, BackwardAPI):
             output_create = ""
 
             for i, out_type_item in enumerate(out_dtype_list):
-                kernel_output = kernel_output + f'kernel_out_{i}, '
+                kernel_output.append(f'kernel_out_{i}')
                 output_names.append(f'kernel_out_{i}')
                 if inplace_flag and self.inplace_map is not None and self.outputs[
                         'names'][i] in self.inplace_map:
@@ -84,7 +84,6 @@ class SparseBackwardAPI(SparseAPI, BackwardAPI):
                 output_create = output_create + f"""
     auto kernel_out_{i} = SetSparseKernelOutput({self.outputs['names'][i]}, {output_type_map[out_dtype_list[i]]});"""
 
-            kernel_output = kernel_output[:-2]
         else:
             raise ValueError(
                 "{} : Output error: the output should not be empty.".format(

paddle/phi/api/yaml/generator/strings_api_gen.py

@@ -55,13 +55,13 @@ class StringsAPI(ForwardAPI):
                     out_tensor_type_list=None,
                     code_indent='',
                     inplace_flag=False):
-        kernel_output = ""
+        kernel_output = []
         output_names = []
         output_create = ""
         return_type = self.get_return_type(inplace_flag)
 
         if len(out_dtype_list) == 1:
-            kernel_output = 'kernel_out'
+            kernel_output.append('kernel_out')
             output_names.append('kernel_out')
             kernel_tensor_out_type = self.get_kernel_tensor_out_type(
                 self.outputs['names'][0])
@@ -78,7 +78,7 @@ class StringsAPI(ForwardAPI):
   {return_type} api_output;"""
 
             for i in range(len(out_dtype_list)):
-                kernel_output = kernel_output + f'kernel_out_{i}, '
+                kernel_output.append(f'kernel_out_{i}')
                 output_names.append(f'kernel_out_{i}')
                 kernel_tensor_out_type = self.get_kernel_tensor_out_type(
                     self.outputs['names'][i])
@@ -91,7 +91,6 @@ class StringsAPI(ForwardAPI):
                 output_create = output_create + f"""
   {tensor_type}* kernel_out_{i} = dynamic_cast<{tensor_type}*>(SetStringsKernelOutput(&std::get<{i}>(api_output), {kernel_tensor_out_type}));"""
 
-            kernel_output = kernel_output[:-2]
         else:
             raise ValueError(
                 "{} : Output error: the output should not be empty.".format(
@@ -178,13 +177,14 @@ class StringsAPI(ForwardAPI):
 
         return f"""
   // 1. Get kernel signature and kernel
-  const auto& kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
-      "{self.kernel['func'][0]}", {{kernel_backend, kernel_layout, kernel_data_type}});
   VLOG(6) << "{self.api} api strings kernel key: [" << kernel_backend << ", " << kernel_layout << ", "<< kernel_data_type << "]";
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+      "{self.kernel['func'][0]}", {{kernel_backend, kernel_layout, kernel_data_type}});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "{self.api} api strings kernel: " << kernel;
 
   // 2. Get Device Context and input
-  auto* dev_ctx = GetDeviceContextByBackend(kernel_backend);
+  auto* dev_ctx = GetDeviceContextByBackend(kernel_result.has_fallback_cpu ? Backend::CPU : kernel_backend);
   {input_tensors}
 
   //  3. Set output
@@ -195,7 +195,7 @@ class StringsAPI(ForwardAPI):
 
 {code_indent}  using kernel_signature = {kernel_signature};
 {code_indent}  auto* kernel_fn = kernel.GetVariadicKernelFn<kernel_signature>();
-{code_indent}  (*kernel_fn)({kernel_args}, {outputs_args});
+{code_indent}  (*kernel_fn)({kernel_args}, {", ".join(outputs_args)});
 
 {code_indent}  {self.gene_return_code()}"""
 

paddle/phi/core/kernel_factory.cc

@@ -84,7 +84,7 @@ bool KernelFactory::HasKernel(const std::string& kernel_name,
   return true;
 }
 
-const Kernel& KernelFactory::SelectKernelOrThrowError(
+KernelResult KernelFactory::SelectKernelOrThrowError(
     const std::string& kernel_name,
     const KernelKey& kernel_key,
     bool use_gpudnn) const {
@@ -104,7 +104,7 @@ const Kernel& KernelFactory::SelectKernelOrThrowError(
           {Backend::GPUDNN, DataLayout::ALL_LAYOUT, kernel_key.dtype()});
     }
     if (kernel_iter != iter->second.end()) {
-      return kernel_iter->second;
+      return {kernel_iter->second, false};
     }
     LOG(WARNING) << "The cudnn kernel for [" << kernel_name
                  << "] is not registered.";
@@ -120,7 +120,7 @@ const Kernel& KernelFactory::SelectKernelOrThrowError(
     kernel_iter = iter->second.find(any_layout_kernel_key);
   }
 
-#ifdef PADDLE_WITH_CUSTOM_DEVICE
+  bool has_fallback_cpu = false;
   if (kernel_iter == iter->second.end()) {
     // Fallback CPU backend
     phi::KernelKey cpu_kernel_key(
@@ -132,8 +132,8 @@ const Kernel& KernelFactory::SelectKernelOrThrowError(
           phi::Backend::CPU, phi::DataLayout::ALL_LAYOUT, kernel_key.dtype());
       kernel_iter = iter->second.find(any_layout_kernel_key);
     }
+    has_fallback_cpu = true;
   }
-#endif
 
   PADDLE_ENFORCE_NE(
       kernel_iter,
@@ -143,16 +143,7 @@ const Kernel& KernelFactory::SelectKernelOrThrowError(
           kernel_key,
           kernel_name));
 
-  return kernel_iter->second;
-}
-
-const Kernel& KernelFactory::SelectKernelOrThrowError(
-    const std::string& kernel_name,
-    Backend backend,
-    DataLayout layout,
-    DataType dtype) const {
-  return SelectKernelOrThrowError(kernel_name,
-                                  KernelKey(backend, layout, dtype));
+  return {kernel_iter->second, has_fallback_cpu};
 }
 
 const KernelArgsDef& KernelFactory::GetFirstKernelArgsDef(

paddle/phi/core/kernel_factory.h

@@ -252,6 +252,14 @@ using KernelKeyMap = paddle::flat_hash_map<KernelKey, Kernel, KernelKey::Hash>;
 
 using KernelNameMap = paddle::flat_hash_map<std::string, KernelKeyMap>;
 
+struct KernelResult {
+  KernelResult(const Kernel& kernel, bool fallback_cpu)
+      : kernel(kernel), has_fallback_cpu(fallback_cpu) {}
+
+  const Kernel& kernel;
+  bool has_fallback_cpu = false;
+};
+
 /**
  * Note: Each Computation need a basic kernel map that named by kernel_name.
  *       Such as for scale op, KernelMap contains a `scale` kernel map,
@@ -268,15 +276,10 @@ class KernelFactory {
     return kernels_.find(TransToPhiKernelName(op_type)) != kernels_.end();
   }
 
-  const Kernel& SelectKernelOrThrowError(const std::string& kernel_name,
+  KernelResult SelectKernelOrThrowError(const std::string& kernel_name,
                                         const KernelKey& kernel_key,
                                         bool use_gpudnn = false) const;
 
-  const Kernel& SelectKernelOrThrowError(const std::string& kernel_name,
-                                         Backend backend,
-                                         DataLayout layout,
-                                         DataType dtype) const;
-
   bool HasKernel(const std::string& kernel_name,
                  const KernelKey& kernel_key) const;
 

paddle/phi/tests/api/scale_api.h

@@ -51,8 +51,9 @@ PADDLE_API Tensor scale_kernel_context(const Tensor& x,
       kernel_data_type = kernel_key.dtype();
     }
   }
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "scale", {kernel_backend, kernel_layout, kernel_data_type});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "scale API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
   VLOG(6) << "scale API kernel: " << kernel;
@@ -221,8 +222,9 @@ Tensor scale_switch_case(const Tensor& x,
       kernel_data_type = kernel_key.dtype();
     }
   }
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       "scale", {kernel_backend, kernel_layout, kernel_data_type});
+  const auto& kernel = kernel_result.kernel;
   VLOG(6) << "scale API kernel key: [" << kernel_backend << ", "
           << kernel_layout << ", " << kernel_data_type << "]";
   VLOG(6) << "scale API kernel: " << kernel;

paddle/phi/tests/core/test_custom_kernel.cc

@@ -192,8 +192,9 @@ TEST(CustomKernel, custom_kernel_dot) {
               fake_dot_kernels.end());
 
   // 4.kernel select
-  auto kernel = phi::KernelFactory::Instance().SelectKernelOrThrowError(
+  auto kernel_result = phi::KernelFactory::Instance().SelectKernelOrThrowError(
       op_name, phi::KernelKey(backend, layout, phi::DataType::UINT8));
+  const auto& kernel = kernel_result.kernel;
 
   // 5.prepare parameters for kernel
   const auto alloc = std::make_unique<paddle::experimental::DefaultAllocator>(