fix encryption and doc

CMakeLists.txt

@@ -58,7 +58,12 @@ option(APP          "Compile Paddle Serving App package"	    OFF)
 option(WITH_ELASTIC_CTR "Compile ELASITC-CTR solution"              OFF)
 option(PACK         "Compile for whl"                               OFF)
 option(WITH_TRT     "Compile Paddle Serving with TRT"       OFF)
+option(PADDLE_ON_INFERENCE "Compile for encryption" ON)
 
+if (PADDLE_ON_INFERENCE)
+    add_definitions(-DPADDLE_ON_INFERENCE)
+    message(STATUS "Use PADDLE_ON_INFERENCE")
+endif()
 set(WITH_MKLML ${WITH_MKL})
 if (NOT DEFINED WITH_MKLDNN)
     if (WITH_MKL AND AVX2_FOUND)

cmake/paddlepaddle.cmake

@@ -124,7 +124,7 @@ LINK_DIRECTORIES(${PADDLE_INSTALL_DIR}/third_party/install/mkldnn/lib)
 ADD_LIBRARY(openblas STATIC IMPORTED GLOBAL)
 SET_PROPERTY(TARGET openblas PROPERTY IMPORTED_LOCATION ${PADDLE_INSTALL_DIR}/third_party/install/openblas/lib/libopenblas.a)
 
-ADD_LIBRARY(paddle_fluid SHARED IMPORTED GLOBAL)
+ADD_LIBRARY(paddle_fluid STATIC IMPORTED GLOBAL)
 SET_PROPERTY(TARGET paddle_fluid PROPERTY IMPORTED_LOCATION ${PADDLE_INSTALL_DIR}/lib/libpaddle_fluid.a)
 
 if (WITH_TRT)

doc/COMPILE.md

@@ -104,7 +104,7 @@ you can execute `make install` to put targets under directory `./output`, you ne
 ### CUDNN_LIBRARY && CUDA_CUDART_LIBRARY is the lib path, it should be /usr/local/cuda/lib64/
  
 ``` shell
-export CUDA_PATH='/usr/local'
+export CUDA_PATH='/usr/local/cuda'
 export CUDNN_LIBRARY='/usr/local/cuda/lib64/'
 export CUDA_CUDART_LIBRARY="/usr/local/cuda/lib64/"
 
@@ -123,7 +123,7 @@ make -j10
 ### Integrated TRT version paddle inference library
 
 ```
-export CUDA_PATH='/usr/local'
+export CUDA_PATH='/usr/local/cuda'
 export CUDNN_LIBRARY='/usr/local/cuda/lib64/'
 export CUDA_CUDART_LIBRARY="/usr/local/cuda/lib64/"
 

doc/COMPILE_CN.md

@@ -100,7 +100,7 @@ make -j10
 ### CUDA_PATH是cuda的安装路径，可以使用命令行whereis cuda命令确认你的cuda安装路径，通常应该是/usr/local/cuda
 ### CUDNN_LIBRARY CUDA_CUDART_LIBRARY 是cuda库文件的路径，通常应该是/usr/local/cuda/lib64/
 ``` shell
-export CUDA_PATH='/usr/local'
+export CUDA_PATH='/usr/local/cuda'
 export CUDNN_LIBRARY='/usr/local/cuda/lib64/'
 export CUDA_CUDART_LIBRARY="/usr/local/cuda/lib64/"
 
@@ -119,7 +119,7 @@ make -j10
 ### 集成TensorRT版本Paddle Inference Library
 
 ```
-export CUDA_PATH='/usr/local'
+export CUDA_PATH='/usr/local/cuda'
 export CUDNN_LIBRARY='/usr/local/cuda/lib64/'
 export CUDA_CUDART_LIBRARY="/usr/local/cuda/lib64/"
 export TENSORRT_LIBRARY_PATH="/usr/local/TensorRT-6.0.1.5/targets/x86_64-linux-gnu/"

paddle_inference/inferencer-fluid-gpu/include/fluid_gpu_engine.h

@@ -283,6 +283,63 @@ class Parameter {
   float* _params;
 };
 
+
+class FluidGpuAnalysisEncryptCore : public FluidFamilyCore {
+ public:
+  void ReadBinaryFile(const std::string& filename, std::string* contents) {
+    std::ifstream fin(filename, std::ios::in | std::ios::binary);
+    fin.seekg(0, std::ios::end);
+    contents->clear();
+    contents->resize(fin.tellg());
+    fin.seekg(0, std::ios::beg);
+    fin.read(&(contents->at(0)), contents->size());
+    fin.close();
+  }
+
+  int create(const predictor::InferEngineCreationParams& params) {
+    std::string data_path = params.get_path();
+    if (access(data_path.c_str(), F_OK) == -1) {
+      LOG(ERROR) << "create paddle predictor failed, path note exits: "
+                 << data_path;
+      return -1;
+    }
+
+    std::string model_buffer, params_buffer, key_buffer;
+    ReadBinaryFile(data_path + "encrypt_model", &model_buffer);
+    ReadBinaryFile(data_path + "encrypt_params", &params_buffer);
+    ReadBinaryFile(data_path + "key", &key_buffer);
+
+    VLOG(2) << "prepare for encryption model";
+
+    auto cipher = paddle::MakeCipher("");
+    std::string real_model_buffer = cipher->Decrypt(model_buffer, key_buffer);
+    std::string real_params_buffer = cipher->Decrypt(params_buffer, key_buffer);
+
+    Config analysis_config;
+    analysis_config.SetModelBuffer(&real_model_buffer[0],
+                                   real_model_buffer.size(),
+                                   &real_params_buffer[0],
+                                   real_params_buffer.size());
+    analysis_config.EnableUseGpu(100, FLAGS_gpuid);
+    analysis_config.SetCpuMathLibraryNumThreads(1);
+    if (params.enable_memory_optimization()) {
+      analysis_config.EnableMemoryOptim();
+    }
+    analysis_config.SwitchSpecifyInputNames(true);
+    AutoLock lock(GlobalPaddleCreateMutex::instance());
+    VLOG(2) << "decrypt model file sucess";
+    _core =
+        CreatePredictor(analysis_config);
+    if (NULL == _core.get()) {
+      LOG(ERROR) << "create paddle predictor failed, path: " << data_path;
+      return -1;
+    }
+    VLOG(2) << "create paddle predictor sucess, path: " << data_path;
+    return 0;
+  }
+};
+
+
 }  // namespace fluid_gpu
 }  // namespace paddle_serving
 }  // namespace baidu

paddle_inference/inferencer-fluid-gpu/src/fluid_gpu_engine.cpp

@@ -31,6 +31,11 @@ REGIST_FACTORY_OBJECT_IMPL_WITH_NAME(
         FluidGpuAnalysisDirCore>,
     ::baidu::paddle_serving::predictor::InferEngine,
     "FLUID_GPU_ANALYSIS_DIR");
+REGIST_FACTORY_OBJECT_IMPL_WITH_NAME(
+    ::baidu::paddle_serving::predictor::FluidInferEngine<
+        FluidGpuAnalysisEncryptCore>,
+    ::baidu::paddle_serving::predictor::InferEngine,
+    "FLUID_GPU_ANALYSIS_ENCRPT")
 
 }  // namespace fluid_gpu
 }  // namespace paddle_serving

python/paddle_serving_server_gpu/serve.py

@@ -21,6 +21,7 @@ import argparse
 import os
 import json
 import base64
+import time
 from multiprocessing import Pool, Process
 from paddle_serving_server_gpu import serve_args
 from flask import Flask, request
@@ -93,6 +94,7 @@ def start_gpu_card_model(index, gpuid, port, args):  # pylint: disable=doc-strin
         server.set_gpuid(gpuid)
     server.run_server()
 
+
 def start_multi_card(args, serving_port=None):  # pylint: disable=doc-string-missing
     gpus = ""
     if serving_port == None:
@@ -121,11 +123,11 @@ def start_multi_card(args, serving_port=None):  # pylint: disable=doc-string-mis
         gpu_processes = []
         for i, gpu_id in enumerate(gpus):
             p = Process(
-		target=start_gpu_card_model,
+                target=start_gpu_card_model,
                 args=(
                     i,
                     gpu_id,
-		    serving_port,
+                    serving_port,
                     args, ))
             gpu_processes.append(p)
         for p in gpu_processes:
@@ -133,6 +135,7 @@ def start_multi_card(args, serving_port=None):  # pylint: disable=doc-string-mis
         for p in gpu_processes:
             p.join()
 
+
 class MainService(BaseHTTPRequestHandler):
     def get_available_port(self):
         default_port = 12000
@@ -227,8 +230,12 @@ if __name__ == "__main__":
         if len(gpu_ids) > 0:
             web_service.set_gpus(gpu_ids)
         web_service.prepare_server(
-            workdir=args.workdir, port=args.port, device=args.device,
-            use_lite=args.use_lite, use_xpu=args.use_xpu, ir_optim=args.ir_optim)
+            workdir=args.workdir,
+            port=args.port,
+            device=args.device,
+            use_lite=args.use_lite,
+            use_xpu=args.use_xpu,
+            ir_optim=args.ir_optim)
         web_service.run_rpc_service()
 
         app_instance = Flask(__name__)

python/paddle_serving_server_gpu/web_service.py

@@ -11,6 +11,7 @@
 # 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.
+#!flask/bin/python
 # pylint: disable=doc-string-missing
 
 from flask import Flask, request, abort
@@ -158,7 +159,7 @@ class WebService(object):
         self.port_list = []
         default_port = 12000
         for i in range(1000):
-	    if port_is_available(default_port + i):
+            if port_is_available(default_port + i):
                 self.port_list.append(default_port + i)
             if len(self.port_list) > len(self.gpus):
                 break