PaddlePaddle · jiangjiajun · Oct 26, 2022 · Oct 23, 2022 · Oct 23, 2022 · Oct 23, 2022
diff --git a/examples/vision/detection/paddledetection/cpp/infer_faster_rcnn.cc b/examples/vision/detection/paddledetection/cpp/infer_faster_rcnn.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::FasterRCNN(
-      model_file, params_file, config_file);
+      model_file, params_file, config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/detection/paddledetection/cpp/infer_mask_rcnn.cc b/examples/vision/detection/paddledetection/cpp/infer_mask_rcnn.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::MaskRCNN(model_file, params_file,
-                                                       config_file);
+                                                       config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/detection/paddledetection/cpp/infer_picodet.cc b/examples/vision/detection/paddledetection/cpp/infer_picodet.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::PicoDet(model_file, params_file,
-                                                      config_file);
+                                                      config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/detection/paddledetection/cpp/infer_ppyolo.cc b/examples/vision/detection/paddledetection/cpp/infer_ppyolo.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::PPYOLO(model_file, params_file,
-                                                     config_file);
+                                                     config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/detection/paddledetection/cpp/infer_ppyoloe.cc b/examples/vision/detection/paddledetection/cpp/infer_ppyoloe.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::PPYOLOE(model_file, params_file,
-                                                      config_file);
+                                                      config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/detection/paddledetection/cpp/infer_yolov3.cc b/examples/vision/detection/paddledetection/cpp/infer_yolov3.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::YOLOv3(model_file, params_file,
-                                                     config_file);
+                                                     config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/detection/paddledetection/cpp/infer_yolox.cc b/examples/vision/detection/paddledetection/cpp/infer_yolox.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::detection::PaddleYOLOX(
-      model_file, params_file, config_file);
+      model_file, params_file, config_file, option);
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/keypointdetection/tiny_pose/cpp/pptinypose_infer.cc b/examples/vision/keypointdetection/tiny_pose/cpp/pptinypose_infer.cc
@@ -25,8 +25,10 @@ void CpuInfer(const std::string& tinypose_model_dir,
   auto tinypose_model_file = tinypose_model_dir + sep + "model.pdmodel";
   auto tinypose_params_file = tinypose_model_dir + sep + "model.pdiparams";
   auto tinypose_config_file = tinypose_model_dir + sep + "infer_cfg.yml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto tinypose_model = fastdeploy::vision::keypointdetection::PPTinyPose(
-      tinypose_model_file, tinypose_params_file, tinypose_config_file);
+      tinypose_model_file, tinypose_params_file, tinypose_config_file, option);
   if (!tinypose_model.Initialized()) {
     std::cerr << "TinyPose Model Failed to initialize." << std::endl;
     return;

diff --git a/examples/vision/matting/ppmatting/cpp/infer.cc b/examples/vision/matting/ppmatting/cpp/infer.cc
@@ -26,6 +26,7 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file,
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "deploy.yaml";
   auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::matting::PPMatting(model_file, params_file,
                                                       config_file, option);
   if (!model.Initialized()) {

diff --git a/examples/vision/segmentation/paddleseg/cpp/infer.cc b/examples/vision/segmentation/paddleseg/cpp/infer.cc
@@ -24,8 +24,10 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file) {
   auto model_file = model_dir + sep + "model.pdmodel";
   auto params_file = model_dir + sep + "model.pdiparams";
   auto config_file = model_dir + sep + "deploy.yaml";
+  auto option = fastdeploy::RuntimeOption();
+  option.UseCpu();
   auto model = fastdeploy::vision::segmentation::PaddleSegModel(
-      model_file, params_file, config_file);
+      model_file, params_file, config_file, option);
 
   if (!model.Initialized()) {
     std::cerr << "Failed to initialize." << std::endl;

diff --git a/fastdeploy/backends/lite/lite_backend.cc b/fastdeploy/backends/lite/lite_backend.cc
@@ -42,7 +42,11 @@ FDDataType LiteDataTypeToFD(const paddle::lite_api::PrecisionType& dtype) {
 void LiteBackend::BuildOption(const LiteBackendOption& option) {
   option_ = option;
   std::vector<paddle::lite_api::Place> valid_places;
-  if (option.enable_fp16) {
+  if (option_.enable_int8) {
+    valid_places.push_back(
+        paddle::lite_api::Place{TARGET(kARM), PRECISION(kInt8)});
+  }
+  if (option_.enable_fp16) {
     paddle::lite_api::MobileConfig check_fp16_config;
     // Determine whether the device supports the FP16
     // instruction set (or whether it is an arm device
@@ -58,12 +62,12 @@ void LiteBackend::BuildOption(const LiteBackendOption& option) {
   valid_places.push_back(
       paddle::lite_api::Place{TARGET(kARM), PRECISION(kFloat)});
   config_.set_valid_places(valid_places);
-  if (option.threads > 0) {
-    config_.set_threads(option.threads);
+  if (option_.threads > 0) {
+    config_.set_threads(option_.threads);
   }
-  if (option.power_mode > 0) {
+  if (option_.power_mode > 0) {
     config_.set_power_mode(
-        static_cast<paddle::lite_api::PowerMode>(option.power_mode));
+        static_cast<paddle::lite_api::PowerMode>(option_.power_mode));
   }
 }
 
@@ -136,14 +140,13 @@ TensorInfo LiteBackend::GetOutputInfo(int index) {
 std::vector<TensorInfo> LiteBackend::GetOutputInfos() { return outputs_desc_; }
 
 bool LiteBackend::Infer(std::vector<FDTensor>& inputs,
-                        std::vector<FDTensor>* outputs) {
+                        std::vector<FDTensor>* outputs) {                                                
   if (inputs.size() != inputs_desc_.size()) {
     FDERROR << "[LiteBackend] Size of inputs(" << inputs.size()
             << ") should keep same with the inputs of this model("
             << inputs_desc_.size() << ")." << std::endl;
     return false;
   }
-
   for (size_t i = 0; i < inputs.size(); ++i) {
     auto iter = inputs_order_.find(inputs[i].name);
     if (iter == inputs_order_.end()) {
@@ -152,12 +155,29 @@ bool LiteBackend::Infer(std::vector<FDTensor>& inputs,
       return false;
     }
     auto tensor = predictor_->GetInput(iter->second);
-    tensor->Resize(inputs[i].shape);
-    tensor->ShareExternalMemory(const_cast<void*>(inputs[i].CpuData()),
-                                inputs[i].Nbytes(),
-                                paddle::lite_api::TargetType::kARM);
+    // Adjust dims only, allocate lazy. 
+    tensor->Resize(inputs[i].shape); 
+    if (inputs[i].dtype == FDDataType::FP32) {
+      tensor->CopyFromCpu<float, paddle::lite_api::TargetType::kARM>(
+        reinterpret_cast<const float*>(const_cast<void*>(
+        inputs[i].CpuData())));
+    } else if (inputs[i].dtype == FDDataType::INT32) {
+      tensor->CopyFromCpu<int, paddle::lite_api::TargetType::kARM>(
+        reinterpret_cast<const int*>(const_cast<void*>(
+        inputs[i].CpuData())));
+    } else if (inputs[i].dtype == FDDataType::INT8) {
+      tensor->CopyFromCpu<int8_t, paddle::lite_api::TargetType::kARM>(
+        reinterpret_cast<const int8_t*>(const_cast<void*>(
+        inputs[i].CpuData())));
+    } else if (inputs[i].dtype == FDDataType::UINT8) {
+      tensor->CopyFromCpu<uint8_t, paddle::lite_api::TargetType::kARM>(
+        reinterpret_cast<const uint8_t*>(const_cast<void*>(
+        inputs[i].CpuData())));
+    } else {
+      FDASSERT(false, "Unexpected data type of %d.", inputs[i].dtype);
+    }
   }
-
+  
   predictor_->Run();
 
   outputs->resize(outputs_desc_.size());

diff --git a/fastdeploy/backends/lite/lite_backend.h b/fastdeploy/backends/lite/lite_backend.h
@@ -37,6 +37,8 @@ struct LiteBackendOption {
   int power_mode = 3;
   // enable fp16
   bool enable_fp16 = false;
+  // enable int8
+  bool enable_int8 = false;
   // optimized model dir for CxxConfig
   std::string optimized_model_dir = "";
   // TODO(qiuyanjun): support more options for lite backend.

diff --git a/fastdeploy/runtime.cc b/fastdeploy/runtime.cc
@@ -317,7 +317,17 @@ void RuntimeOption::EnableLiteFP16() {
   lite_enable_fp16 = true;
 }
 
-void RuntimeOption::DisableLiteFP16() { lite_enable_fp16 = false; }
+void RuntimeOption::DisableLiteFP16() { 
+  lite_enable_fp16 = false; 
+}
+
+void RuntimeOption::EnableLiteInt8() {
+  lite_enable_int8 = true;
+}
+
+void RuntimeOption::DisableLiteInt8() { 
+  lite_enable_int8 = false; 
+}
 
 void RuntimeOption::SetLitePowerMode(LitePowerMode mode) {
   lite_power_mode = mode;
@@ -643,6 +653,7 @@ void Runtime::CreateLiteBackend() {
 #ifdef ENABLE_LITE_BACKEND
   auto lite_option = LiteBackendOption();
   lite_option.threads = option.cpu_thread_num;
+  lite_option.enable_int8 = option.lite_enable_int8;
   lite_option.enable_fp16 = option.lite_enable_fp16;
   lite_option.power_mode = static_cast<int>(option.lite_power_mode);
   lite_option.optimized_model_dir = option.lite_optimized_model_dir;

diff --git a/fastdeploy/runtime.h b/fastdeploy/runtime.h
@@ -171,6 +171,16 @@ struct FASTDEPLOY_DECL RuntimeOption {
    */
   void DisableLiteFP16();
 
+   /**
+   * @brief enable int8 precision while use paddle lite backend
+   */
+  void EnableLiteInt8();
+
+  /**
+   * @brief disable int8 precision, change to full precision(float32)
+   */
+  void DisableLiteInt8();
+
   /**
    * @brief Set power mode while using Paddle Lite as inference backend, mode(0: LITE_POWER_HIGH; 1: LITE_POWER_LOW; 2: LITE_POWER_FULL; 3: LITE_POWER_NO_BIND, 4: LITE_POWER_RAND_HIGH; 5: LITE_POWER_RAND_LOW, refer [paddle lite](https://paddle-lite.readthedocs.io/zh/latest/api_reference/cxx_api_doc.html#set-power-mode) for more details)
    */
@@ -256,6 +266,8 @@ struct FASTDEPLOY_DECL RuntimeOption {
   // 3: LITE_POWER_NO_BIND 4: LITE_POWER_RAND_HIGH
   // 5: LITE_POWER_RAND_LOW
   LitePowerMode lite_power_mode = LitePowerMode::LITE_POWER_NO_BIND;
+  // enable int8 or not
+  bool lite_enable_int8 = false;
   // enable fp16 or not
   bool lite_enable_fp16 = false;
   // optimized model dir for CxxConfig

diff --git a/fastdeploy/vision/detection/ppdet/mask_rcnn.cc b/fastdeploy/vision/detection/ppdet/mask_rcnn.cc
@@ -24,7 +24,7 @@ MaskRCNN::MaskRCNN(const std::string& model_file,
                    const RuntimeOption& custom_option,
                    const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::PDINFER};
+  valid_cpu_backends = {Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::PDINFER};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;

diff --git a/fastdeploy/vision/detection/ppdet/ppyolo.cc b/fastdeploy/vision/detection/ppdet/ppyolo.cc
@@ -23,7 +23,7 @@ PPYOLO::PPYOLO(const std::string& model_file, const std::string& params_file,
                const RuntimeOption& custom_option,
                const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::OPENVINO, Backend::PDINFER};
+  valid_cpu_backends = {Backend::OPENVINO, Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::PDINFER};
   has_nms_ = true;
   runtime_option = custom_option;

diff --git a/fastdeploy/vision/detection/ppdet/ppyoloe.cc b/fastdeploy/vision/detection/ppdet/ppyoloe.cc
@@ -14,7 +14,7 @@ PPYOLOE::PPYOLOE(const std::string& model_file, const std::string& params_file,
                  const RuntimeOption& custom_option,
                  const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::OPENVINO, Backend::ORT, Backend::PDINFER};
+  valid_cpu_backends = {Backend::OPENVINO, Backend::ORT, Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::ORT, Backend::PDINFER, Backend::TRT};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;

diff --git a/fastdeploy/vision/detection/ppdet/rcnn.cc b/fastdeploy/vision/detection/ppdet/rcnn.cc
@@ -24,7 +24,7 @@ FasterRCNN::FasterRCNN(const std::string& model_file,
                        const RuntimeOption& custom_option,
                        const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::PDINFER};
+  valid_cpu_backends = {Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::PDINFER};
   has_nms_ = true;
   runtime_option = custom_option;

diff --git a/fastdeploy/vision/detection/ppdet/yolov3.cc b/fastdeploy/vision/detection/ppdet/yolov3.cc
@@ -23,7 +23,7 @@ YOLOv3::YOLOv3(const std::string& model_file, const std::string& params_file,
                const RuntimeOption& custom_option,
                const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::OPENVINO, Backend::ORT, Backend::PDINFER};
+  valid_cpu_backends = {Backend::OPENVINO, Backend::ORT, Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::ORT, Backend::PDINFER, Backend::TRT};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;

diff --git a/fastdeploy/vision/detection/ppdet/yolox.cc b/fastdeploy/vision/detection/ppdet/yolox.cc
@@ -24,7 +24,7 @@ PaddleYOLOX::PaddleYOLOX(const std::string& model_file,
                          const RuntimeOption& custom_option,
                          const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::ORT, Backend::PDINFER};
+  valid_cpu_backends = {Backend::ORT, Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::ORT, Backend::PDINFER, Backend::TRT};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;

diff --git a/fastdeploy/vision/faceid/contrib/insightface_rec.cc b/fastdeploy/vision/faceid/contrib/insightface_rec.cc
@@ -30,7 +30,7 @@ InsightFaceRecognitionModel::InsightFaceRecognitionModel(
     valid_cpu_backends = {Backend::ORT};
     valid_gpu_backends = {Backend::ORT, Backend::TRT};
   } else {
-    valid_cpu_backends = {Backend::PDINFER, Backend::ORT};
+    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::LITE};
     valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
   }
   runtime_option = custom_option;

diff --git a/fastdeploy/vision/keypointdet/pptinypose/pptinypose.cc b/fastdeploy/vision/keypointdet/pptinypose/pptinypose.cc
@@ -16,7 +16,7 @@ PPTinyPose::PPTinyPose(const std::string& model_file,
                        const RuntimeOption& custom_option,
                        const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO};
+  valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO, Backend::LITE};
   valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;

diff --git a/fastdeploy/vision/matting/ppmatting/ppmatting.cc b/fastdeploy/vision/matting/ppmatting/ppmatting.cc
@@ -26,7 +26,7 @@ PPMatting::PPMatting(const std::string& model_file,
                      const RuntimeOption& custom_option,
                      const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::ORT, Backend::PDINFER};
+  valid_cpu_backends = {Backend::ORT, Backend::PDINFER, Backend::LITE};
   valid_gpu_backends = {Backend::PDINFER, Backend::TRT};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;

diff --git a/fastdeploy/vision/ocr/ppocr/classifier.cc b/fastdeploy/vision/ocr/ppocr/classifier.cc
@@ -30,7 +30,7 @@ Classifier::Classifier(const std::string& model_file,
                           Backend::OPENVINO}; 
     valid_gpu_backends = {Backend::ORT, Backend::TRT};  
   } else {
-    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO};
+    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO, Backend::LITE};
     valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
   }
   runtime_option = custom_option;

diff --git a/fastdeploy/vision/ocr/ppocr/dbdetector.cc b/fastdeploy/vision/ocr/ppocr/dbdetector.cc
@@ -30,7 +30,7 @@ DBDetector::DBDetector(const std::string& model_file,
                           Backend::OPENVINO};  
     valid_gpu_backends = {Backend::ORT, Backend::TRT};  
   } else {
-    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO};
+    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO, Backend::LITE};
     valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
   }
 

diff --git a/fastdeploy/vision/ocr/ppocr/ppocr_v2.cc b/fastdeploy/vision/ocr/ppocr/ppocr_v2.cc
@@ -110,7 +110,6 @@ bool PPOCRv2::Predict(cv::Mat* img,
     if (nullptr != classifier_ && result->cls_labels[i] % 2 == 1 && result->cls_scores[i] > classifier_->cls_thresh) {
       cv::rotate(image_list[i], image_list[i], 1);
     }
-
     if (nullptr != recognizer_ && !Recognize(&(image_list[i]), result)) {
       FDERROR << "Failed to recgnize croped image of index " << i << "." << std::endl;
       return false;

diff --git a/fastdeploy/vision/ocr/ppocr/recognizer.cc b/fastdeploy/vision/ocr/ppocr/recognizer.cc
@@ -48,7 +48,7 @@ Recognizer::Recognizer(const std::string& model_file,
                           Backend::OPENVINO};  
     valid_gpu_backends = {Backend::ORT, Backend::TRT};  
   } else {
-    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO};
+    valid_cpu_backends = {Backend::PDINFER, Backend::ORT, Backend::OPENVINO, Backend::LITE};
     valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
   }
 

diff --git a/fastdeploy/vision/segmentation/ppseg/model.cc b/fastdeploy/vision/segmentation/ppseg/model.cc
@@ -27,7 +27,7 @@ PaddleSegModel::PaddleSegModel(const std::string& model_file,
                                const RuntimeOption& custom_option,
                                const ModelFormat& model_format) {
   config_file_ = config_file;
-  valid_cpu_backends = {Backend::OPENVINO, Backend::PDINFER, Backend::ORT};
+  valid_cpu_backends = {Backend::OPENVINO, Backend::PDINFER, Backend::ORT, Backend::LITE};
   valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
   runtime_option = custom_option;
   runtime_option.model_format = model_format;
@@ -107,7 +107,7 @@ bool PaddleSegModel::BuildPreprocessPipelineFromConfig() {
 	        << "Please refer to https://github.com/PaddlePaddle/PaddleSeg/blob/develop/docs/model_export.md"
 	        << " to export model with fixed input shape."
 	        << std::endl;
-      valid_cpu_backends = {Backend::OPENVINO, Backend::PDINFER};
+      valid_cpu_backends = {Backend::OPENVINO, Backend::PDINFER, Backend::LITE};
       valid_gpu_backends = {Backend::PDINFER};
     }
     if (input_height != -1 && input_width != -1 && !yml_contain_resize_op) {