[Protobuf] add combined-param model save/load supported test=develop (#1876)

lite/api/apis_test.cc

@@ -57,8 +57,11 @@ TEST(CXXApi_LightApi, optim_model) {
   });
   // On ARM devices, the preferred X86 target not works, but it can still
   // select ARM kernels.
-  cxx_api.Build(
-      FLAGS_model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places);
+  cxx_api.Build(FLAGS_model_dir,
+                "",
+                "",
+                Place{TARGET(kX86), PRECISION(kFloat)},
+                valid_places);
   cxx_api.SaveModel(FLAGS_optimized_model);
 }
 
@@ -75,8 +78,11 @@ TEST(CXXApi_LightApi, save_and_load_model) {
     });
     // On ARM devices, the preferred X86 target not works, but it can still
     // select ARM kernels.
-    cxx_api.Build(
-        FLAGS_model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places);
+    cxx_api.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kX86), PRECISION(kFloat)},
+                  valid_places);
 
     auto* x = cxx_api.GetInput(0);
     SetConstInput(x);

lite/api/cxx_api.cc

@@ -33,7 +33,7 @@ void Predictor::SaveModel(const std::string &dir,
   program_->SaveOpInfosToProgram(&program_desc_);
   switch (model_type) {
     case lite_api::LiteModelType::kProtobuf:
-      SaveModelPb(dir, *program_->exec_scope(), program_desc_);
+      SaveModelPb(dir, *program_->exec_scope(), program_desc_, true);
       break;
     case lite_api::LiteModelType::kNaiveBuffer:
       SaveModelNaive(dir, *program_->exec_scope(), program_desc_);
@@ -84,16 +84,29 @@ const cpp::ProgramDesc &Predictor::program_desc() const {
 const RuntimeProgram &Predictor::runtime_program() const { return *program_; }
 
 void Predictor::Build(const std::string &model_path,
+                      const std::string model_file,
+                      const std::string param_file,
                       const Place &prefer_place,
                       const std::vector<Place> &valid_places,
                       const std::vector<std::string> &passes,
                       lite_api::LiteModelType model_type) {
   LOG(INFO) << "Load model from " << model_path;
   switch (model_type) {
-    case lite_api::LiteModelType::kProtobuf:
-      LoadModelPb(model_path, scope_.get(), &program_desc_);
-      break;
+    case lite_api::LiteModelType::kProtobuf: {
+      bool combined_param = false;
+      if (!model_file.empty() && !param_file.empty()) {
+        combined_param = true;
+      }
+      LoadModelPb(model_path,
+                  model_file,
+                  param_file,
+                  scope_.get(),
+                  &program_desc_,
+                  combined_param);
+    } break;
     case lite_api::LiteModelType::kNaiveBuffer:
+      CHECK(!model_path.empty())
+          << "NaiveBuffer backend only supported combined param";
       LoadModelNaive(model_path, scope_.get(), &program_desc_);
       break;
     default:

lite/api/cxx_api.h

@@ -41,6 +41,8 @@ class LITE_API Predictor {
   // Build from a model, with places set for hardware config.
   void Build(
       const std::string& model_path,
+      const std::string model_file_path,
+      const std::string param_file_path,
       const Place& prefer_place,
       const std::vector<Place>& valid_places,
       const std::vector<std::string>& passes = {},

lite/api/cxx_api_bin.cc

@@ -41,7 +41,7 @@ void Run(const char* model_dir, int repeat) {
   });
 
   predictor.Build(
-      model_dir, Place{TARGET(kARM), PRECISION(kInt8)}, valid_places);
+      model_dir, "", "", Place{TARGET(kARM), PRECISION(kInt8)}, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/cxx_api_impl.cc

@@ -47,7 +47,11 @@ CxxPaddleApiImpl::CxxPaddleApiImpl() {}
 void CxxPaddleApiImpl::Init(const lite_api::CxxConfig &config) {
   auto places = config.valid_places();
   places.emplace_back(TARGET(kHost), PRECISION(kAny), DATALAYOUT(kAny));
-  raw_predictor_.Build(config.model_dir(), config.preferred_place(), places);
+  raw_predictor_.Build(config.model_dir(),
+                       config.model_file(),
+                       config.param_file(),
+                       config.preferred_place(),
+                       places);
 }
 
 std::unique_ptr<lite_api::Tensor> CxxPaddleApiImpl::GetInput(int i) {

lite/api/cxx_api_test.cc

@@ -45,8 +45,11 @@ TEST(CXXApi, save_model) {
   lite::Predictor predictor;
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kX86), PRECISION(kFloat)}});
-  predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kCUDA), PRECISION(kFloat)}, valid_places);
+  predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kCUDA), PRECISION(kFloat)},
+                  valid_places);
 
   LOG(INFO) << "Save optimized model to " << FLAGS_optimized_model;
   predictor.SaveModel(FLAGS_optimized_model,
@@ -93,8 +96,11 @@ TEST(CXXApi, save_model) {
   lite::Predictor predictor;
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kARM), PRECISION(kFloat)}});
-  predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kARM), PRECISION(kFloat)}, valid_places);
+  predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kARM), PRECISION(kFloat)},
+                  valid_places);
 
   LOG(INFO) << "Save optimized model to " << FLAGS_optimized_model;
   predictor.SaveModel(FLAGS_optimized_model);
@@ -107,6 +113,8 @@ TEST(CXXApi, load_model_naive) {
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kARM), PRECISION(kFloat)}});
   predictor.Build(FLAGS_optimized_model + ".naive",
+                  "",
+                  "",
                   Place{TARGET(kARM), PRECISION(kFloat)},
                   valid_places,
                   {},

lite/api/efficientnet_b0_test.cc

@@ -31,7 +31,7 @@ void TestModel(const std::vector<Place> &valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto *input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/inceptionv4_test.cc

@@ -33,8 +33,11 @@ TEST(InceptionV4, test) {
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kARM), PRECISION(kFloat)}});
 
-  predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kARM), PRECISION(kFloat)}, valid_places);
+  predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kARM), PRECISION(kFloat)},
+                  valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/light_api.cc

@@ -24,7 +24,7 @@ void LightPredictor::Build(const std::string& model_dir,
   switch (model_type) {
 #ifndef LITE_ON_TINY_PUBLISH
     case lite_api::LiteModelType::kProtobuf:
-      LoadModelPb(model_dir, scope_.get(), &desc);
+      LoadModelPb(model_dir, "", "", scope_.get(), &desc);
       break;
 #endif
     case lite_api::LiteModelType::kNaiveBuffer:

lite/api/lite_api_test_helper.cc

@@ -38,6 +38,8 @@ const lite::Tensor* RunHvyModel() {
 #endif
 
   predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
                   Place{TARGET(kX86), PRECISION(kFloat)},  // origin cuda
                   valid_places);
 

lite/api/mobilenetv1_int8_test.cc

@@ -32,7 +32,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/mobilenetv1_ssd_test.cc

@@ -32,7 +32,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 300, 300})));

lite/api/mobilenetv1_test.cc

@@ -36,7 +36,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(model_dir, preferred_place, valid_places);
+  predictor.Build(model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/mobilenetv1_yolov3_test.cc

@@ -32,7 +32,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 608, 608})));

lite/api/mobilenetv2_test.cc

@@ -37,7 +37,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(model_dir, preferred_place, valid_places);
+  predictor.Build(model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/model_optimize_tool.cc

@@ -23,7 +23,12 @@
 #include "lite/utils/cp_logging.h"
 #include "lite/utils/string.h"
 
-DEFINE_string(model_dir, "", "path of the model");
+DEFINE_string(model_dir,
+              "",
+              "path of the model. This option will be ignored if model_file "
+              "and param_file are exist");
+DEFINE_string(model_file, "", "model file path of the combined-param model");
+DEFINE_string(param_file, "", "param file path of the combined-param model");
 DEFINE_string(
     optimize_out_type,
     "protobuf",
@@ -39,8 +44,15 @@ namespace paddle {
 namespace lite_api {
 
 void Main() {
+  if (!FLAGS_model_file.empty() && !FLAGS_param_file.empty()) {
+    LOG(WARNING)
+        << "Load combined-param model. Option model_dir will be ignored";
+  }
+
   lite_api::CxxConfig config;
   config.set_model_dir(FLAGS_model_dir);
+  config.set_model_file(FLAGS_model_file);
+  config.set_param_file(FLAGS_param_file);
 
   std::vector<Place> valid_places;
   auto target_reprs = lite::Split(FLAGS_valid_targets, " ");

lite/api/model_run_test_image.cc

@@ -39,7 +39,7 @@ TEST(model, test) {
     precision = PRECISION(kInt8);
   }
   predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kARM), precision}, valid_places);
+      FLAGS_model_dir, "", "", Place{TARGET(kARM), precision}, valid_places);
   int im_width = FLAGS_im_width;
   int im_height = FLAGS_im_height;
   auto* input_tensor = predictor.GetInput(0);

lite/api/ocr_attention_test.cc

@@ -32,7 +32,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 1, 48, 512})));

lite/api/paddle_api.h

@@ -99,13 +99,19 @@ class LITE_API ConfigBase {
 class LITE_API CxxConfig : public ConfigBase {
   Place preferred_place_;
   std::vector<Place> valid_places_;
+  std::string model_file_;
+  std::string param_file_;
 
  public:
   void set_preferred_place(const Place& x) { preferred_place_ = x; }
   void set_valid_places(const std::vector<Place>& x) { valid_places_ = x; }
+  void set_model_file(const std::string& path) { model_file_ = path; }
+  void set_param_file(const std::string& path) { param_file_ = path; }
 
   const Place& preferred_place() const { return preferred_place_; }
   const std::vector<Place>& valid_places() const { return valid_places_; }
+  std::string model_file() const { return model_file_; }
+  std::string param_file() const { return param_file_; }
 };
 
 /// MobileConfig is the config for the light weight predictor, it will skip

lite/api/resnet18_test.cc

@@ -31,8 +31,11 @@ TEST(ResNet18, test) {
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kARM), PRECISION(kFloat)}});
 
-  predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kARM), PRECISION(kFloat)}, valid_places);
+  predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kARM), PRECISION(kFloat)},
+                  valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/resnet50_test.cc

@@ -32,7 +32,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/resnet50_test_fpga.cc

@@ -33,6 +33,8 @@ TEST(ResNet50, test) {
        Place{TARGET(kHost), PRECISION(kFloat), DATALAYOUT(kNHWC)}});
 
   predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
                   Place{TARGET(kFPGA), PRECISION(kFP16), DATALAYOUT(kNHWC)},
                   valid_places);
 

lite/api/shufflenetv2_test.cc

@@ -31,7 +31,7 @@ void TestModel(const std::vector<Place>& valid_places,
   DeviceInfo::Global().SetRunMode(lite_api::LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
 
-  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
+  predictor.Build(FLAGS_model_dir, "", "", preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim((std::vector<DDim::value_type>({1, 3, 224, 224}))));

lite/api/test_googlenet_lite.cc

@@ -51,7 +51,7 @@ TEST(CXXApi, test_lite_googlenet) {
   //  LOG(INFO)<<"FLAGS_eval_googlenet_dir:"<<FLAGS_test_lite_googlenet_dir;
   std::string model_dir = FLAGS_model_dir;
   predictor.Build(
-      model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places);
+      model_dir, "", "", Place{TARGET(kX86), PRECISION(kFloat)}, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/test_inceptionv4_lite_x86.cc

@@ -55,8 +55,12 @@ TEST(InceptionV4, test_inceptionv4_lite_x86) {
                                    "io_copy_kernel_pick_pass",
                                    "variable_place_inference_pass",
                                    "runtime_context_assign_pass"});
-  predictor.Build(
-      model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places, passes);
+  predictor.Build(model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kX86), PRECISION(kFloat)},
+                  valid_places,
+                  passes);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/test_mobilenetv1_lite_x86.cc

@@ -54,8 +54,12 @@ TEST(Mobilenet_v1, test_mobilenetv1_lite_x86) {
                                    "io_copy_kernel_pick_pass",
                                    "variable_place_inference_pass",
                                    "runtime_context_assign_pass"});
-  predictor.Build(
-      model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places, passes);
+  predictor.Build(model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kX86), PRECISION(kFloat)},
+                  valid_places,
+                  passes);
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));
   auto* data = input_tensor->mutable_data<float>();

lite/api/test_mobilenetv2_lite_x86.cc

@@ -56,8 +56,12 @@ TEST(Mobilenet_v2, test_mobilenetv2_lite_x86) {
                                    "io_copy_kernel_pick_pass",
                                    "variable_place_inference_pass",
                                    "runtime_context_assign_pass"});
-  predictor.Build(
-      model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places, passes);
+  predictor.Build(model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kX86), PRECISION(kFloat)},
+                  valid_places,
+                  passes);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));

lite/api/unet_test.cc

@@ -33,8 +33,11 @@ TEST(unet, test) {
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kARM), PRECISION(kFloat)}});
 
-  predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kARM), PRECISION(kFloat)}, valid_places);
+  predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kARM), PRECISION(kFloat)},
+                  valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 512, 512})));

lite/core/mir/fusion/fc_fuse_pass_test.cc

@@ -44,6 +44,8 @@ TEST(fc_fuse_pass, fuse_test) {
 #endif
 
   predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
                   Place{TARGET(kX86), PRECISION(kFloat)},  // origin cuda
                   valid_places);
 
@@ -72,8 +74,11 @@ TEST(fc_fuse_pass, save_model_test) {
   lite::Predictor predictor;
   std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
                                    Place{TARGET(kX86), PRECISION(kFloat)}});
-  predictor.Build(
-      FLAGS_model_dir, Place{TARGET(kX86), PRECISION(kFloat)}, valid_places);
+  predictor.Build(FLAGS_model_dir,
+                  "",
+                  "",
+                  Place{TARGET(kX86), PRECISION(kFloat)},
+                  valid_places);
 
   LOG(INFO) << "Save optimized model to " << FLAGS_optimized_model;
   predictor.SaveModel(FLAGS_optimized_model);

lite/core/mir/subgraph/generate_npu_program_pass_test.cc

@@ -40,7 +40,7 @@ void TestModel(lite::Predictor* predictor,
                const std::vector<Place>& valid_places,
                const std::string& model_dir) {
   predictor->Build(
-      model_dir, Place{TARGET(kARM), PRECISION(kFloat)}, valid_places);
+      model_dir, "", "", Place{TARGET(kARM), PRECISION(kFloat)}, valid_places);
 
   auto* input_tensor = predictor->GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>(

lite/core/mir/subgraph/subgraph_program_pass_test.cc

@@ -32,7 +32,7 @@ namespace lite {
 TEST(SubgraphTest, mobilenetv2) {
   cpp::ProgramDesc program_desc;
   auto scope = std::make_shared<Scope>();
-  LoadModelPb(FLAGS_model_dir, scope.get(), &program_desc);
+  LoadModelPb(FLAGS_model_dir, "", "", scope.get(), &program_desc);
   std::vector<Place> valid_places({
       Place{TARGET(kHost), PRECISION(kFloat)},
 #ifdef LITE_WITH_ARM

lite/gen_code/gen_code_test.cc

@@ -145,7 +145,10 @@ TEST(gen_code, auto_gen) {
 TEST(gen_code, optimized_program) {
   lite::Scope scope;
   cpp::ProgramDesc cpp_desc;
-  LoadModelPb(FLAGS_optimized_model, &scope, &cpp_desc);
+  std::string model_file = FLAGS_optimized_model + "/model";
+  std::string param_file = FLAGS_optimized_model + "/params";
+  LoadModelPb(
+      FLAGS_optimized_model, model_file, param_file, &scope, &cpp_desc, true);
 
   framework::proto::ProgramDesc pb_proto_desc;
   lite::pb::ProgramDesc pb_desc(&pb_proto_desc);

lite/gen_code/paddle_code_generator.cc

@@ -27,7 +27,9 @@ namespace gencode {
 void GenCode(const std::string& model_dir, const std::string& out_file) {
   lite::Scope scope;
   cpp::ProgramDesc cpp_desc;
-  LoadModelPb(model_dir, &scope, &cpp_desc);
+  std::string model_file = model_dir + "/model";
+  std::string param_file = model_dir + "/params";
+  LoadModelPb(model_dir, model_file, param_file, &scope, &cpp_desc, true);
 
   framework::proto::ProgramDesc pb_proto_desc;
   lite::pb::ProgramDesc pb_desc(&pb_proto_desc);

lite/model_parser/model_parser.cc

@@ -159,15 +159,63 @@ void LoadParam(const std::string &path, Variable *out) {
   LoadLoDTensor(fin, out);
 }
 
+//
+bool IsPersistable(const cpp::VarDesc &var) {
+  if (var.Persistable() && var.GetType() != VarDescAPI::Type::FEED_MINIBATCH &&
+      var.GetType() != VarDescAPI::Type::FETCH_LIST &&
+      var.GetType() != VarDescAPI::Type::RAW) {
+    return true;
+  }
+  return false;
+}
+
+void LoadCombinedParamsPb(const std::string &path,
+                          lite::Scope *scope,
+                          const cpp::ProgramDesc &cpp_prog) {
+  CHECK(scope);
+  auto prog = cpp_prog;
+  auto &main_block_desc = *prog.GetBlock<cpp::BlockDesc>(0);
+
+  // Get vars
+  std::vector<std::string> paramlist;
+  for (size_t i = 0; i < main_block_desc.VarsSize(); ++i) {
+    auto &var = *main_block_desc.GetVar<cpp::VarDesc>(i);
+    if (!IsPersistable(var)) continue;
+    paramlist.push_back(var.Name());
+  }
+  std::sort(paramlist.begin(), paramlist.end());
+
+  // Load vars
+  std::ifstream file(path);
+  CHECK(file.is_open());
+  for (size_t i = 0; i < paramlist.size(); ++i) {
+    auto *var = scope->Var(paramlist[i]);
+    // Error checking
+    CHECK(static_cast<bool>(file))
+        << "There is a problem with loading model parameters";
+    LoadLoDTensor(file, var);
+  }
+  file.peek();
+  CHECK(file.eof()) << "You are not allowed to load partial data via"
+                    << " LoadCombinedParamsPb, use LoadParam instead.";
+  file.close();
+}
+
 void LoadModelPb(const std::string &model_dir,
+                 const std::string &model_file,
+                 const std::string &param_file,
                  Scope *scope,
-                 cpp::ProgramDesc *cpp_prog) {
+                 cpp::ProgramDesc *cpp_prog,
+                 bool combined) {
   CHECK(cpp_prog);
   CHECK(scope);
   cpp_prog->ClearBlocks();
 
   // Load model
-  const std::string prog_path = model_dir + "/__model__";
+  std::string prog_path = model_dir + "/__model__";
+  if (combined) {
+    prog_path = model_file;
+  }
   framework::proto::ProgramDesc pb_proto_prog = *LoadProgram(prog_path);
   pb::ProgramDesc pb_prog(&pb_proto_prog);
 
@@ -176,6 +224,9 @@ void LoadModelPb(const std::string &model_dir,
 
   // Load Params
   // NOTE: Only main block be used now.
+  if (combined) {
+    LoadCombinedParamsPb(param_file, scope, *cpp_prog);
+  } else {
     auto main_block = pb_proto_prog.blocks(0);
     for (auto &var : main_block.vars()) {
       if (var.name() == "feed" || var.name() == "fetch" || !var.persistable())
@@ -193,6 +244,8 @@ void LoadModelPb(const std::string &model_dir,
           CHECK(false) << "unknown weight type";
       }
     }
+  }
+
 #ifdef LITE_WITH_NPU
   for (auto &op : main_block.ops()) {
     LOG(INFO) << "op type:" << op.type();
@@ -216,14 +269,18 @@ void LoadModelPb(const std::string &model_dir,
 
 void SaveModelPb(const std::string &model_dir,
                  const Scope &exec_scope,
-                 const cpp::ProgramDesc &cpp_prog) {
+                 const cpp::ProgramDesc &cpp_prog,
+                 bool combined) {
   MkDirRecur(model_dir);
   // Save program
   framework::proto::ProgramDesc pb_proto_prog;
   pb::ProgramDesc pb_prog(&pb_proto_prog);
   TransformProgramDescCppToAny(cpp_prog, &pb_prog);
 
-  const std::string prog_path = model_dir + "/__model__";
+  std::string prog_path = model_dir + "/__model__";
+  if (combined) {
+    prog_path = model_dir + "/model";
+  }
   std::ofstream model_ostream(prog_path, std::ios_base::binary);
   CHECK(model_ostream.is_open());
   const std::string pb_str = pb_proto_prog.SerializeAsString();
@@ -232,8 +289,13 @@ void SaveModelPb(const std::string &model_dir,
 
   // Save Params
   // NOTE: Only main block be used now.
+  if (combined) {
+    const std::string combined_params_path = model_dir + "/params";
+    SaveCombinedParamsPb(combined_params_path, exec_scope, cpp_prog);
+  } else {
     for (auto &item : pb_proto_prog.blocks(0).vars()) {
-    if (item.name() == "feed" || item.name() == "fetch" || !item.persistable())
+      if (item.name() == "feed" || item.name() == "fetch" ||
+          !item.persistable())
         continue;
       const std::string path = model_dir + "/" + item.name();
       std::ofstream var_ostream(path, std::ios::binary);
@@ -241,9 +303,34 @@ void SaveModelPb(const std::string &model_dir,
       SerializeTensor(var_ostream, exec_scope, item.name());
       var_ostream.close();
     }
+  }
   VLOG(4) << "Save protobuf model in '" << model_dir << "'' successfully";
 }
 
+void SaveCombinedParamsPb(const std::string &path,
+                          const lite::Scope &exec_scope,
+                          const cpp::ProgramDesc &cpp_prog) {
+  auto prog = cpp_prog;
+  auto &main_block_desc = *prog.GetBlock<cpp::BlockDesc>(0);
+
+  // Get vars
+  std::vector<std::string> paramlist;
+  for (size_t i = 0; i < main_block_desc.VarsSize(); ++i) {
+    auto &var = *main_block_desc.GetVar<cpp::VarDesc>(i);
+    if (!IsPersistable(var)) continue;
+    paramlist.push_back(var.Name());
+  }
+  std::sort(paramlist.begin(), paramlist.end());
+
+  // Load vars
+  std::ofstream file(path);
+  CHECK(file.is_open());
+  for (size_t i = 0; i < paramlist.size(); ++i) {
+    SerializeTensor(file, exec_scope, paramlist[i]);
+  }
+  file.close();
+}
+
 void TensorToStream(std::ostream &os, const lite::Tensor &tensor) {
   // the 1st field, uint32_t version
   constexpr uint32_t version = 0;

lite/model_parser/model_parser.h

@@ -41,14 +41,26 @@ void LoadParams(const std::string& path);
 // Load a single parameter to an output tensor.
 void LoadParam(const std::string& path, Variable* out);
 
+void LoadCombinedParamsPb(const std::string& path,
+                          lite::Scope* scope,
+                          const cpp::ProgramDesc& prog);
+
 // Read a model and files of parameters in pb format.
 void LoadModelPb(const std::string& model_dir,
+                 const std::string& model_file,
+                 const std::string& param_file,
                  Scope* scope,
-                 cpp::ProgramDesc* prog);
+                 cpp::ProgramDesc* prog,
+                 bool combined = false);
 
 // Save a model and files of parameters in pb format.
 void SaveModelPb(const std::string& model_dir,
                  const Scope& scope,
+                 const cpp::ProgramDesc& prog,
+                 bool combined = false);
+
+void SaveCombinedParamsPb(const std::string& path,
+                          const lite::Scope& exec_scope,
                           const cpp::ProgramDesc& prog);
 
 // Serialize tensors to ostream.

lite/model_parser/model_parser_test.cc

@@ -40,18 +40,38 @@ TEST(ModelParser, LoadModelPb) {
   CHECK(!FLAGS_model_dir.empty());
   cpp::ProgramDesc prog;
   Scope scope;
-  LoadModelPb(FLAGS_model_dir, &scope, &prog);
+  LoadModelPb(FLAGS_model_dir, "", "", &scope, &prog);
 }
 
 TEST(ModelParser, SaveModelPb) {
   CHECK(!FLAGS_model_dir.empty());
   cpp::ProgramDesc prog;
   Scope scope;
-  LoadModelPb(FLAGS_model_dir, &scope, &prog);
+  LoadModelPb(FLAGS_model_dir, "", "", &scope, &prog);
   const std::string save_pb_model_path = FLAGS_model_dir + ".saved.pb";
   SaveModelPb(save_pb_model_path, scope, prog);
 }
 
+TEST(ModelParser, SaveModelCombinedPb) {
+  CHECK(!FLAGS_model_dir.empty());
+  cpp::ProgramDesc prog;
+  Scope scope;
+  LoadModelPb(FLAGS_model_dir, "", "", &scope, &prog);
+  const std::string save_pb_model_path = FLAGS_model_dir + ".saved.pb.combined";
+  SaveModelPb(save_pb_model_path, scope, prog, true);
+}
+
+TEST(ModelParser, LoadModelCombinedPb) {
+  CHECK(!FLAGS_model_dir.empty());
+  const std::string model_path = FLAGS_model_dir + ".saved.pb.combined";
+  cpp::ProgramDesc prog;
+  Scope scope;
+  std::string model_file_path = FLAGS_model_dir + ".saved.pb.combined/model";
+  std::string param_file_path = FLAGS_model_dir + ".saved.pb.combined/params";
+  LoadModelPb(
+      model_path, model_file_path, param_file_path, &scope, &prog, true);
+}
+
 TEST(ModelParser, SaveParamNaive) {
   Scope scope;
   auto* tensor = scope.Var("xxx")->GetMutable<lite::Tensor>();
@@ -94,7 +114,7 @@ TEST(ModelParser, SaveModelNaive) {
   CHECK(!FLAGS_model_dir.empty());
   cpp::ProgramDesc prog;
   Scope scope;
-  LoadModelPb(FLAGS_model_dir, &scope, &prog);
+  LoadModelPb(FLAGS_model_dir, "", "", &scope, &prog);
   const std::string save_pb_model_path = FLAGS_model_dir + ".saved.naive";
   SaveModelNaive(save_pb_model_path, scope, prog);
 }

lite/tools/debug/model_debug_tool.cc

@@ -61,6 +61,8 @@ void Run(DebugConfig* conf) {
   }};
 
   predictor.Build(conf->model_dir,
+                  "",
+                  "",
 #ifdef LITE_WITH_ARM
                   Place{TARGET(kARM), PRECISION(kFloat)},
 #endif