Merge branch 'polish_prepare' into add_kl_quant

test_tipc/docs/test_serving_infer_cpp.md

@@ -40,20 +40,40 @@ Linux GPU/CPU C++ 服务化部署测试的主程序为`test_serving_infer_cpp.sh
 
 
 - 安装PaddlePaddle：如果您已经安装了2.2或者以上版本的paddlepaddle，那么无需运行下面的命令安装paddlepaddle。
-    ```shell
-    # 需要安装2.2及以上版本的Paddle
-    # 安装GPU版本的Paddle
-    python3.7 -m pip install paddlepaddle-gpu==2.2.0
-    # 安装CPU版本的Paddle
-    python3.7 -m pip install paddlepaddle==2.2.0
-    ```
+  ```shell
+  # 需要安装2.2及以上版本的Paddle
+  # 安装GPU版本的Paddle
+  python3.7 -m pip install paddlepaddle-gpu==2.2.0
+  # 安装CPU版本的Paddle
+  python3.7 -m pip install paddlepaddle==2.2.0
+  ```
 
 - 安装依赖
-    ```shell
-    python3.7 -m pip install  -r requirements.txt
-    ```
+  ```shell
+  python3.7 -m pip install -r requirements.txt
+  ```
+
+- 安装TensorRT
+  编译 serving-server 的脚本内会设置 `TENSORRT_LIBRARY_PATH` 这一环境变量，因此编译前需要安装TensorRT。
+
+  如果使用`registry.baidubce.com/paddlepaddle/paddle:latest-dev-cuda10.1-cudnn7-gcc82`镜像进测试，则已自带TensorRT无需安装，
+  否则可以参考 [3.2 安装TensorRT](install.md#32-安装tensorrt) 进行安装，并在修改 [build_server.sh](../../deploy/paddleserving/build_server.sh#L62) 的 `TENSORRT_LIBRARY_PATH` 地址为安装后的路径。
+
 - 安装 PaddleServing 相关组件，包括serving_client、serving-app，自动编译并安装带自定义OP的 serving_server 包，以及自动下载并解压推理模型
   ```bash
+  # 安装必要依赖包
+  python3.7 -m pip install paddle_serving_client==0.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
+  python3.7 -m pip install paddle-serving-app==0.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+  # 安装编译自定义OP的serving-server包
+  pushd ./deploy/paddleserving
+  source build_server.sh python3.7
+  popd
+
+  # 测试PP-ShiTu识别模型时需安装faiss包
+  python3.7-m pip install faiss-cpu==1.7.1post2 -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+  # 下载模型与数据
   bash test_tipc/prepare.sh test_tipc/configs/PPLCNet/PPLCNet_x1_0_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt serving_infer
   ```
 
@@ -62,14 +82,14 @@ Linux GPU/CPU C++ 服务化部署测试的主程序为`test_serving_infer_cpp.sh
 测试方法如下所示，希望测试不同的模型文件，只需更换为自己的参数配置文件，即可完成对应模型的测试。
 
 ```bash
-bash test_tipc/test_serving_infer_cpp.sh ${your_params_file}
+bash test_tipc/test_serving_infer_cpp.sh ${your_params_file} ${mode}
 ```
 
 以`PPLCNet_x1_0`的`Linux GPU/CPU C++ 服务化部署测试`为例，命令如下所示。
 
 
 ```bash
-bash test_tipc/test_serving_infer_cpp.sh test_tipc/configs/PPLCNet/PPLCNet_x1_0_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
+bash test_tipc/test_serving_infer_cpp.sh test_tipc/configs/PPLCNet/PPLCNet_x1_0_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt serving_infer
 ```
 
 输出结果如下，表示命令运行成功。

test_tipc/docs/test_serving_infer_python.md

@@ -40,20 +40,30 @@ Linux GPU/CPU PYTHON 服务化部署测试的主程序为`test_serving_infer_pyt
 
 
 - 安装PaddlePaddle：如果您已经安装了2.2或者以上版本的paddlepaddle，那么无需运行下面的命令安装paddlepaddle。
-    ```shell
-    # 需要安装2.2及以上版本的Paddle
-    # 安装GPU版本的Paddle
-    python3.7 -m pip install paddlepaddle-gpu==2.2.0
-    # 安装CPU版本的Paddle
-    python3.7 -m pip install paddlepaddle==2.2.0
-    ```
+  ```shell
+  # 需要安装2.2及以上版本的Paddle
+  # 安装GPU版本的Paddle
+  python3.7 -m pip install paddlepaddle-gpu==2.2.0
+  # 安装CPU版本的Paddle
+  python3.7 -m pip install paddlepaddle==2.2.0
+  ```
 
 - 安装依赖
-    ```shell
-    python3.7 -m pip install  -r requirements.txt
-    ```
+  ```shell
+  python3.7 -m pip install -r requirements.txt
+  ```
+
 - 安装 PaddleServing 相关组件，包括serving-server、serving_client、serving-app，自动下载并解压推理模型
   ```bash
+  # 安装必要依赖包
+  python3.7 -m pip install paddle_serving_client==0.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
+  python3.7 -m pip install paddle-serving-app==0.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
+  python3.7 -m pip install install paddle-serving-server-gpu==0.9.0.post101 -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+  # 测试PP-ShiTu识别模型时需安装faiss包
+  python3.7-m pip install faiss-cpu==1.7.1post2 -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+  # 下载模型与数据
   bash test_tipc/prepare.sh test_tipc/configs/ResNet50/ResNet50_linux_gpu_normal_normal_serving_python_linux_gpu_cpu.txt serving_infer
   ```
 
@@ -62,14 +72,14 @@ Linux GPU/CPU PYTHON 服务化部署测试的主程序为`test_serving_infer_pyt
 测试方法如下所示，希望测试不同的模型文件，只需更换为自己的参数配置文件，即可完成对应模型的测试。
 
 ```bash
-bash test_tipc/test_serving_infer_python.sh ${your_params_file}
+bash test_tipc/test_serving_infer_python.sh ${your_params_file} ${mode}
 ```
 
 以`ResNet50`的`Linux GPU/CPU PYTHON 服务化部署测试`为例，命令如下所示。
 
 
 ```bash
-bash test_tipc/test_serving_infer_python.sh test_tipc/configs/ResNet50/ResNet50_linux_gpu_normal_normal_serving_python_linux_gpu_cpu.txt
+bash test_tipc/test_serving_infer_python.sh test_tipc/configs/ResNet50/ResNet50_linux_gpu_normal_normal_serving_python_linux_gpu_cpu.txt serving_infer
 ```
 
 输出结果如下，表示命令运行成功。

test_tipc/prepare.sh

@@ -206,18 +206,7 @@ fi
 if [[ ${MODE} = "serving_infer" ]]; then
     # prepare serving env
     python_name=$(func_parser_value "${lines[2]}")
-    ${python_name} -m pip install paddle_serving_client==0.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
-    ${python_name} -m pip install paddle-serving-app==0.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
-    python_name=$(func_parser_value "${lines[2]}")
-    if [[ ${FILENAME} =~ "cpp" ]]; then
-        pushd ./deploy/paddleserving
-        source build_server.sh ${python_name}
-        popd
-    else
-        ${python_name} -m pip install install paddle-serving-server-gpu==0.9.0.post101 -i https://pypi.tuna.tsinghua.edu.cn/simple
-    fi
     if [[ ${model_name} =~ "ShiTu" ]]; then
-        ${python_name} -m pip install faiss-cpu==1.7.1post2 -i https://pypi.tuna.tsinghua.edu.cn/simple
         cls_inference_model_url=$(func_parser_value "${lines[3]}")
         cls_tar_name=$(func_get_url_file_name "${cls_inference_model_url}")
         det_inference_model_url=$(func_parser_value "${lines[4]}")

test_tipc/test_serving_infer_cpp.sh

@@ -104,6 +104,8 @@ function func_serving_cls(){
         if [[ ${use_gpu} = "null" ]]; then
             web_service_cpp_cmd="${python_} -m paddle_serving_server.serve --model ${serving_server_dir_name} --op GeneralClasOp --port 9292 &"
             eval ${web_service_cpp_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cpp_cmd}" "${status_log}" "${model_name}"
             sleep 5s
             _save_log_path="${LOG_PATH}/server_infer_cpp_cpu_pipeline_batchsize_1.log"
             pipeline_cmd="${python_} test_cpp_serving_client.py > ${_save_log_path} 2>&1 "
@@ -116,6 +118,8 @@ function func_serving_cls(){
         else
             web_service_cpp_cmd="${python_} -m paddle_serving_server.serve --model ${serving_server_dir_name} --op GeneralClasOp --port 9292 --gpu_id=${use_gpu} &"
             eval ${web_service_cpp_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cpp_cmd}" "${status_log}" "${model_name}"
             sleep 8s
 
             _save_log_path="${LOG_PATH}/server_infer_cpp_gpu_pipeline_batchsize_1.log"
@@ -207,12 +211,14 @@ function func_serving_rec(){
     unset https_proxy
     unset http_proxy
 
-    export SERVING_BIN=${PWD}/../Serving/server-build-gpu-opencv/core/general-server/serving
+    # export SERVING_BIN=${PWD}/../Serving/server-build-gpu-opencv/core/general-server/serving
     for use_gpu in ${web_use_gpu_list[*]}; do
         if [ ${use_gpu} = "null" ]; then
             det_serving_server_dir_name=$(func_get_url_file_name "$det_serving_server_value")
             web_service_cpp_cmd="${python_interp} -m paddle_serving_server.serve --model ../../${det_serving_server_value} ../../${cls_serving_server_value} --op GeneralPicodetOp GeneralFeatureExtractOp --port 9400 &"
             eval ${web_service_cpp_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cpp_cmd}" "${status_log}" "${model_name}"
             sleep 5s
             _save_log_path="${LOG_PATH}/server_infer_cpp_cpu_batchsize_1.log"
             pipeline_cmd="${python_interp} ${pipeline_py} > ${_save_log_path} 2>&1 "
@@ -226,6 +232,8 @@ function func_serving_rec(){
             det_serving_server_dir_name=$(func_get_url_file_name "$det_serving_server_value")
             web_service_cpp_cmd="${python_interp} -m paddle_serving_server.serve --model ../../${det_serving_server_value} ../../${cls_serving_server_value} --op GeneralPicodetOp GeneralFeatureExtractOp --port 9400 --gpu_id=${use_gpu} &"
             eval ${web_service_cpp_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cpp_cmd}" "${status_log}" "${model_name}"
             sleep 5s
             _save_log_path="${LOG_PATH}/server_infer_cpp_gpu_batchsize_1.log"
             pipeline_cmd="${python_interp} ${pipeline_py} > ${_save_log_path} 2>&1 "
@@ -241,9 +249,9 @@ function func_serving_rec(){
 
 
 # set cuda device
-GPUID=$2
+GPUID=$3
 if [ ${#GPUID} -le 0 ];then
-    env=" "
+    env="export CUDA_VISIBLE_DEVICES=0"
 else
     env="export CUDA_VISIBLE_DEVICES=${GPUID}"
 fi

test_tipc/test_serving_infer_python.sh

@@ -98,6 +98,8 @@ function func_serving_cls(){
 
             web_service_cmd="${python_} ${web_service_py} &"
             eval ${web_service_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cmd}" "${status_log}" "${model_name}"
             sleep 5s
             for pipeline in ${pipeline_py[*]}; do
                 _save_log_path="${LOG_PATH}/server_infer_cpu_${pipeline%_client*}_batchsize_1.log"
@@ -130,6 +132,8 @@ function func_serving_cls(){
 
             web_service_cmd="${python_} ${web_service_py} & "
             eval ${web_service_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cmd}" "${status_log}" "${model_name}"
             sleep 5s
             for pipeline in ${pipeline_py[*]}; do
                 _save_log_path="${LOG_PATH}/server_infer_gpu_${pipeline%_client*}_batchsize_1.log"
@@ -237,6 +241,8 @@ function func_serving_rec(){
 
             web_service_cmd="${python} ${web_service_py} &"
             eval ${web_service_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cmd}" "${status_log}" "${model_name}"
             sleep 5s
             for pipeline in ${pipeline_py[*]}; do
                 _save_log_path="${LOG_PATH}/server_infer_cpu_${pipeline%_client*}_batchsize_1.log"
@@ -269,6 +275,8 @@ function func_serving_rec(){
 
             web_service_cmd="${python} ${web_service_py} & "
             eval ${web_service_cmd}
+            last_status=${PIPESTATUS[0]}
+            status_check $last_status "${web_service_cmd}" "${status_log}" "${model_name}"
             sleep 10s
             for pipeline in ${pipeline_py[*]}; do
                 _save_log_path="${LOG_PATH}/server_infer_gpu_${pipeline%_client*}_batchsize_1.log"
@@ -288,9 +296,9 @@ function func_serving_rec(){
 
 
 # set cuda device
-GPUID=$2
+GPUID=$3
 if [ ${#GPUID} -le 0 ];then
-    env=" "
+    env="export CUDA_VISIBLE_DEVICES=0"
 else
     env="export CUDA_VISIBLE_DEVICES=${GPUID}"
 fi