[tipc] fix test_train_infer_python/cpp (#5516)

* add pact about * fix doc * fix comments * add ptq aware * fix * add benchmark.png * fix png and ptq * fix tipc * fix train_infer_python * delete mkldnn trt * fix cpp * fix comment * fix comment * fix * status_check * fix comment * add build cpp to prepare.sh
PaddlePaddle · May 18, 2022 · 34f046e · 34f046e
1 parent 2aa2d44
commit 34f046e
Show file tree

Hide file tree

Showing 19 changed files with 120 additions and 291 deletions.
diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/README.md b/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/README.md
@@ -91,7 +91,7 @@ cd build
 cmake  .. \
     -DWITH_CONTRIB=OFF \
     -DWITH_MKL=ON \
-    -DWITH_MKLDNN=ON  \
+    -DWITH_MKLDNN=OFF  \
     -DWITH_TESTING=OFF \
     -DCMAKE_BUILD_TYPE=Release \
     -DWITH_INFERENCE_API_TEST=OFF \
@@ -205,9 +205,6 @@ make -j
   * gpu_id：使用的GPU卡号；
   * gpu_mem：显存；
   * cpu_math_library_num_threads：底层科学计算库所用线程的数量；
-  * use_mkldnn：是否使用MKLDNN加速；
-  * use_tensorrt: 是否使用tensorRT进行加速；
-  * use_fp16：是否使用半精度浮点数进行计算，该选项仅在use_tensorrt为true时有效；
   * cls_model_path：预测模型结构文件路径；
   * cls_params_path：预测模型参数文件路径；
   * resize_short_size：预处理时图像缩放大小；

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/include/cls.h b/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/include/cls.h
@@ -40,16 +40,12 @@ class Classifier {
                       const std::string &params_path, const bool &use_gpu,
                       const int &gpu_id, const int &gpu_mem,
                       const int &cpu_math_library_num_threads,
-                      const bool &use_mkldnn, const bool &use_tensorrt,
-                      const bool &use_fp16, const int &resize_short_size,
+                      const int &resize_short_size,
                       const int &crop_size) {
     this->use_gpu_ = use_gpu;
     this->gpu_id_ = gpu_id;
     this->gpu_mem_ = gpu_mem;
     this->cpu_math_library_num_threads_ = cpu_math_library_num_threads;
-    this->use_mkldnn_ = use_mkldnn;
-    this->use_tensorrt_ = use_tensorrt;
-    this->use_fp16_ = use_fp16;
 
     this->resize_short_size_ = resize_short_size;
     this->crop_size_ = crop_size;
@@ -70,9 +66,6 @@ class Classifier {
   int gpu_id_ = 0;
   int gpu_mem_ = 4000;
   int cpu_math_library_num_threads_ = 4;
-  bool use_mkldnn_ = false;
-  bool use_tensorrt_ = false;
-  bool use_fp16_ = false;
 
   std::vector<float> mean_ = {0.485f, 0.456f, 0.406f};
   std::vector<float> scale_ = {1 / 0.229f, 1 / 0.224f, 1 / 0.225f};

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/include/cls_config.h b/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/include/cls_config.h
@@ -39,11 +39,6 @@ class ClsConfig {
     this->cpu_math_library_num_threads =
         stoi(config_map_["cpu_math_library_num_threads"]);
 
-    this->use_mkldnn = bool(stoi(config_map_["use_mkldnn"]));
-
-    this->use_tensorrt = bool(stoi(config_map_["use_tensorrt"]));
-    this->use_fp16 = bool(stoi(config_map_["use_fp16"]));
-
     this->cls_model_path.assign(config_map_["cls_model_path"]);
 
     this->cls_params_path.assign(config_map_["cls_params_path"]);
@@ -61,11 +56,6 @@ class ClsConfig {
 
   int cpu_math_library_num_threads = 1;
 
-  bool use_mkldnn = false;
-
-  bool use_tensorrt = false;
-  bool use_fp16 = false;
-
   std::string cls_model_path;
 
   std::string cls_params_path;

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/src/cls.cpp b/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/src/cls.cpp
@@ -24,20 +24,8 @@ void Classifier::LoadModel(const std::string &model_path,
 
   if (this->use_gpu_) {
     config.EnableUseGpu(this->gpu_mem_, this->gpu_id_);
-    if (this->use_tensorrt_) {
-      config.EnableTensorRtEngine(
-          1 << 20, 1, 3,
-          this->use_fp16_ ? paddle_infer::Config::Precision::kHalf
-                          : paddle_infer::Config::Precision::kFloat32,
-          false, false);
-    }
   } else {
     config.DisableGpu();
-    if (this->use_mkldnn_) {
-      config.EnableMKLDNN();
-      // cache 10 different shapes for mkldnn to avoid memory leak
-      config.SetMkldnnCacheCapacity(10);
-    }
     config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_);
   }
 

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/src/main.cpp b/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/src/main.cpp
@@ -61,8 +61,7 @@ int main(int argc, char **argv) {
 
   Classifier classifier(config.cls_model_path, config.cls_params_path,
                         config.use_gpu, config.gpu_id, config.gpu_mem,
-                        config.cpu_math_library_num_threads, config.use_mkldnn,
-                        config.use_tensorrt, config.use_fp16,
+                        config.cpu_math_library_num_threads,
                         config.resize_short_size, config.crop_size);
 
   double elapsed_time = 0.0;

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/tools/build.sh b/tutorials/mobilenetv3_prod/Step6/deploy/inference_cpp/tools/build.sh
@@ -2,7 +2,6 @@ OPENCV_DIR=../opencv-3.4.7/opencv3/
 LIB_DIR=../paddle_inference/
 CUDA_LIB_DIR=/usr/local/cuda/lib64
 CUDNN_LIB_DIR=/usr/lib64
-TENSORRT_DIR=/usr/local/TensorRT-7.2.3.4
 
 BUILD_DIR=build
 rm -rf ${BUILD_DIR}

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_python/README.md b/tutorials/mobilenetv3_prod/Step6/deploy/inference_python/README.md
@@ -13,7 +13,7 @@
 
 ## 1. 简介
 
-Paddle Inference 是飞桨的原生推理库， 作用于服务器端和云端，提供高性能的推理能力。相比于直接基于预训练模型进行预测，Paddle Inference可使用MKLDNN、CUDNN、TensorRT进行预测加速，从而实现更优的推理性能。
+Paddle Inference 是飞桨的原生推理库， 作用于服务器端和云端，提供高性能的推理能力。
 
 本文档主要基于Paddle Inference的mobilenet_v3_small模型推理。
 

diff --git a/tutorials/mobilenetv3_prod/Step6/deploy/inference_python/infer.py b/tutorials/mobilenetv3_prod/Step6/deploy/inference_python/infer.py
@@ -80,6 +80,8 @@ def load_predictor(self, model_file_path, params_file_path):
             config.enable_use_gpu(1000, 0)
         else:
             config.disable_gpu()
+            # The thread num should not be greater than the number of cores in the CPU.
+            config.set_cpu_math_library_num_threads(4)
 
         # enable memory optim
         config.enable_memory_optim()

diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/README.md b/tutorials/mobilenetv3_prod/Step6/test_tipc/README.md
@@ -16,6 +16,7 @@
 - Slim训练部署：包括PACT在线量化、离线量化。
 - 更多训练环境：包括Windows GPU/CPU、Linux NPU、Linux DCU等多种环境。
 
+
 | 算法论文 | 模型名称 | 模型类型 | 基础<br>训练预测 | 更多<br>训练方式 | 更多<br>部署方式 | Slim<br>训练部署 |  更多<br>训练环境  |
 | :--- | :--- |  :----:  | :--------: |  :----:  |   :----:  |   :----:  |   :----:  |
 | MobileNetV3     | mobilenet_v3_small |  分类  | 支持 | 混合精度 | PYTHON 服务化部署<br>Paddle2ONNX 部署| PACT量化<br>离线量化 | Windows GPU/CPU |

diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/common_func.sh b/tutorials/mobilenetv3_prod/Step6/test_tipc/common_func.sh
@@ -65,10 +65,11 @@ function status_check(){
     last_status=$1   # the exit code
     run_command=$2
     run_log=$3
+    model_name=$4
     if [ $last_status -eq 0 ]; then
-        echo -e "\033[33m Run successfully with command - ${run_command}!  \033[0m" | tee -a ${run_log}
+        echo -e "\033[33m Run successfully with command - ${model_name} - ${run_command}!  \033[0m" | tee -a ${run_log}
     else
-        echo -e "\033[33m Run failed with command - ${run_command}!  \033[0m" | tee -a ${run_log}
+        echo -e "\033[33m Run failed with command - ${model_name} - ${run_command}!  \033[0m" | tee -a ${run_log}
     fi
 }
 
diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/configs/mobilenet_v3_small/inference_cpp.txt b/tutorials/mobilenetv3_prod/Step6/test_tipc/configs/mobilenet_v3_small/inference_cpp.txt
@@ -1,14 +1,12 @@
 # model load config
+model_name mobilenet_v3_small
 use_gpu  0
 gpu_id  0
 gpu_mem  4000
 cpu_math_library_num_threads  10
-use_mkldnn 1
-use_tensorrt 0
-use_fp16 0
 
 # cls config
 cls_model_path  ./deploy/inference_cpp/mobilenet_v3_small_infer/inference.pdmodel
 cls_params_path ./deploy/inference_cpp/mobilenet_v3_small_infer/inference.pdiparams
 resize_short_size 256
-crop_size 224
+crop_size 224
diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/prepare.sh b/tutorials/mobilenetv3_prod/Step6/test_tipc/prepare.sh
@@ -54,6 +54,47 @@ elif [ ${MODE} = "serving_infer" ];then
         wget -nc -P  ./inference https://paddle-model-ecology.bj.bcebos.com/model/mobilenetv3_reprod/mobilenet_v3_small_infer.tar  --no-check-certificate
         cd ./inference && tar xf mobilenet_v3_small_infer.tar && cd ../
     fi
+elif [ ${MODE} = "cpp_infer" ];then
+    # wget model
+    wget -nc -P  ./deploy/inference_cpp/ https://paddle-model-ecology.bj.bcebos.com/model/mobilenetv3_reprod/mobilenet_v3_small_infer.tar  --no-check-certificate
+    cd ./deploy/inference_cpp/ && tar xf mobilenet_v3_small_infer.tar
+    wget -nc https://paddle-inference-lib.bj.bcebos.com/2.2.2/cxx_c/Linux/GPU/x86-64_gcc8.2_avx_mkl_cuda11.1_cudnn8.1.1_trt7.2.3.4/paddle_inference.tgz
+    tar zxf paddle_inference.tgz
+    wget -nc https://paddleocr.bj.bcebos.com/libs/opencv/opencv-3.4.7.tar.gz
+    tar zxf opencv-3.4.7.tar.gz
+    # build opencv
+    cd opencv-3.4.7/
+    root_path=$PWD
+    install_path=${root_path}/opencv3
+    build_dir=${root_path}/build
+
+    rm -rf ${build_dir}
+    mkdir ${build_dir}
+    cd ${build_dir}
+
+    cmake .. \
+    -DCMAKE_INSTALL_PREFIX=${install_path} \
+    -DCMAKE_BUILD_TYPE=Release \
+    -DBUILD_SHARED_LIBS=OFF \
+    -DWITH_IPP=OFF \
+    -DBUILD_IPP_IW=OFF \
+    -DWITH_LAPACK=OFF \
+    -DWITH_EIGEN=OFF \
+    -DCMAKE_INSTALL_LIBDIR=lib64 \
+    -DWITH_ZLIB=ON \
+    -DBUILD_ZLIB=ON \
+    -DWITH_JPEG=ON \
+    -DBUILD_JPEG=ON \
+    -DWITH_PNG=ON \
+    -DBUILD_PNG=ON \
+    -DWITH_TIFF=ON \
+    -DBUILD_TIFF=ON
+    make -j
+    make install
+    cd ../../
+    # build cpp
+    bash tools/build.sh
+
 elif [ ${MODE} = "paddle2onnx_infer" ];then
     # get data
     tar -xf ./test_images/lite_data.tar

diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/test_inference_cpp.sh b/tutorials/mobilenetv3_prod/Step6/test_tipc/test_inference_cpp.sh
@@ -28,45 +28,28 @@ IFS=$'\n'
 lines=(${dataline})
 
 # parser load config
-use_gpu_key=$(func_parser_key_cpp "${lines[1]}")
-use_gpu_value=$(func_parser_value_cpp "${lines[1]}")
-use_mkldnn_key=$(func_parser_key_cpp "${lines[5]}")
-use_mkldnn_value=$(func_parser_value_cpp "${lines[5]}")
-use_tensorrt_key=$(func_parser_key_cpp "${lines[6]}")
-use_tensorrt_value=$(func_parser_value_cpp "${lines[6]}")
-use_fp16_key=$(func_parser_key_cpp "${lines[7]}")
-use_fp16_value=$(func_parser_value_cpp "${lines[7]}")
+model_name=$(func_parser_value_cpp "${lines[1]}")
+use_gpu_key=$(func_parser_key_cpp "${lines[2]}")
+use_gpu_value=$(func_parser_value_cpp "${lines[2]}")
 
 LOG_PATH="./log/infer_cpp"
 mkdir -p ${LOG_PATH}
 status_log="${LOG_PATH}/results_infer_cpp.log"
 
 function func_infer_cpp(){
     # inference cpp
-    if test $use_gpu_value -gt 0; then
-        if test $use_tensorrt_value -gt 0; then
-            if test $use_fp16_value -gt 0; then
-                _save_log_path="${LOG_PATH}/infer_cpp_${use_gpu_key}_${use_tensorrt_key}_${use_fp16_key}.log"
-            else
-                _save_log_path="${LOG_PATH}/infer_cpp_${use_gpu_key}_${use_tensorrt_key}.log"
-            fi
-        else
-            _save_log_path="${LOG_PATH}/infer_cpp_${use_gpu_key}.log"
-        fi
+    if test $use_gpu_value -gt 0; then     
+        _save_log_path="${LOG_PATH}/infer_cpp_use_cpu.log" 
     else
-        if test $use_mkldnn_value -gt 0; then
-            _save_log_path="${LOG_PATH}/infer_cpp_use_cpu_${use_mkldnn_key}.log"
-        else
-            _save_log_path="${LOG_PATH}/infer_cpp_use_cpu.log"
-        fi    
+        _save_log_path="${LOG_PATH}/infer_cpp_${use_gpu_key}.log"
     fi
     # run infer cpp
     inference_cpp_cmd="./deploy/inference_cpp/build/clas_system"
     inference_cpp_img="./images/demo.jpg" 
     infer_cpp_full_cmd="${inference_cpp_cmd} ${FILENAME} ${inference_cpp_img} > ${_save_log_path} 2>&1 "   
     eval $infer_cpp_full_cmd
     last_status=${PIPESTATUS[0]}
-    status_check $last_status "${infer_cpp_full_cmd}" "${status_log}"
+    status_check $last_status "${infer_cpp_full_cmd}" "${status_log}"  "${model_name}"
 }
 
 echo "################### run test cpp inference ###################"

diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/test_ptq_inference_python.sh b/tutorials/mobilenetv3_prod/Step6/test_tipc/test_ptq_inference_python.sh
@@ -65,7 +65,7 @@ function func_inference(){
                 eval $command
                 last_status=${PIPESTATUS[0]}
                 eval "cat ${_save_log_path}"
-                status_check $last_status "${command}" "${status_log}"
+                status_check $last_status "${command}" "${status_log}" "${model_name}"
             done
         # gpu        
         elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then
@@ -80,7 +80,7 @@ function func_inference(){
                 eval $command
                 last_status=${PIPESTATUS[0]}
                 eval "cat ${_save_log_path}"
-                status_check $last_status "${command}" "${status_log}"   
+                status_check $last_status "${command}" "${status_log}"  "${model_name}" 
             done      
         else
             echo "Does not support hardware other than CPU and GPU Currently!"
@@ -107,7 +107,7 @@ if [ ${MODE} = "whole_infer" ]; then
     echo $export_cmd
     # eval $export_cmd
     status_export=$?
-    status_check $status_export "${export_cmd}" "${status_log}"
+    status_check $status_export "${export_cmd}" "${status_log}" "${model_name}"
 
     save_infer_dir=${output_dir_value}
     #run inference

diff --git a/tutorials/mobilenetv3_prod/Step6/test_tipc/test_train_inference_python.sh b/tutorials/mobilenetv3_prod/Step6/test_tipc/test_train_inference_python.sh
@@ -87,7 +87,7 @@ function func_inference(){
                 eval $command
                 last_status=${PIPESTATUS[0]}
                 eval "cat ${_save_log_path}"
-                status_check $last_status "${command}" "${status_log}"
+                status_check $last_status "${command}" "${status_log}" "${model_name}"
             done
         # gpu        
         elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then
@@ -101,7 +101,7 @@ function func_inference(){
                 eval $command
                 last_status=${PIPESTATUS[0]}
                 eval "cat ${_save_log_path}"
-                status_check $last_status "${command}" "${status_log}"   
+                status_check $last_status "${command}" "${status_log}"  "${model_name}"
             done      
         else
             echo "Does not support hardware other than CPU and GPU Currently!"
@@ -131,7 +131,7 @@ if [ ${MODE} = "whole_infer" ]; then
         echo $export_cmd
         eval $export_cmd
         status_export=$?
-        status_check $status_export "${export_cmd}" "${status_log}"
+        status_check $status_export "${export_cmd}" "${status_log}" "${model_name}"
     else
         save_infer_dir=${save_infer_dir}
     fi
@@ -196,7 +196,7 @@ else
             fi
             # run train
             eval $cmd
-            status_check $? "${cmd}" "${status_log}"
+            status_check $? "${cmd}" "${status_log}" "${model_name}"
             # TODO ensure model name
             set_eval_pretrain=$(func_set_params "${pretrain_model_key}" "${save_log}/${train_model_name}")
 
@@ -205,7 +205,7 @@ else
                 eval ${env}
                 eval_cmd="${python} ${eval_py} ${set_eval_pretrain}" 
                 eval $eval_cmd
-                status_check $? "${eval_cmd}" "${status_log}"
+                status_check $? "${eval_cmd}" "${status_log}" "${model_name}"
             fi
             # run export model
             if [ ${run_export} != "null" ]; then 
@@ -215,7 +215,7 @@ else
                 set_save_infer_key=$(func_set_params "${save_infer_key}" "${save_infer_path}")
                 export_cmd="${python} ${run_export} ${set_export_weight} ${set_save_infer_key}"
                 eval $export_cmd
-                status_check $? "${export_cmd}" "${status_log}"
+                status_check $? "${export_cmd}" "${status_log}" "${model_name}"
 
                 #run inference
                 eval $env

diff --git a/tutorials/tipc/infer_cpp/infer_cpp.md b/tutorials/tipc/infer_cpp/infer_cpp.md
@@ -71,20 +71,8 @@ void Classifier::LoadModel(const std::string &model_path,
 
   if (this->use_gpu_) {
     config.EnableUseGpu(this->gpu_mem_, this->gpu_id_);
-    if (this->use_tensorrt_) {
-      config.EnableTensorRtEngine(
-          1 << 20, 1, 3,
-          this->use_fp16_ ? paddle_infer::Config::Precision::kHalf
-                          : paddle_infer::Config::Precision::kFloat32,
-          false, false);
-    }
   } else {
     config.DisableGpu();
-    if (this->use_mkldnn_) {
-      config.EnableMKLDNN();
-      // cache 10 different shapes for mkldnn to avoid memory leak
-      config.SetMkldnnCacheCapacity(10);
-    }
     config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_);
   }