[TIPC] add serving cpp test (PaddlePaddle#2180)

2742195759 · Jun 7, 2022 · 9dca662 · 9dca662
1 parent 79bfcc4
commit 9dca662
Show file tree

Hide file tree

Showing 28 changed files with 895 additions and 37 deletions.
diff --git a/...c/configs/deeplabv3p_resnet50/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/...c/configs/deeplabv3p_resnet50/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:pp_humanseg_server
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/pp_humanseg_server/pp_humanseg_server_export_512x512
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/...plabv3p_resnet50/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...plabv3p_resnet50/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:pp_humanseg_server
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/pp_humanseg_server/pp_humanseg_server_export_512x512
 --model_filename:model.pdmodel

diff --git a/test_tipc/configs/fcn_hrnetw18/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/test_tipc/configs/fcn_hrnetw18/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:fcn_hrnetw18
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/fcn_hrnetw18/fcn_hrnetw18_cityscapes_1024x512_80k
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/...igs/fcn_hrnetw18/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...igs/fcn_hrnetw18/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:fcn_hrnetw18
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/fcn_hrnetw18/fcn_hrnetw18_cityscapes_1024x512_80k
 --model_filename:model.pdmodel

diff --git a/...pc/configs/fcn_hrnetw18_small/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/...pc/configs/fcn_hrnetw18_small/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:pp_humanseg_mobile
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/pp_humanseg_mobile/pp_humanseg_mobile_export_192x192
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/...n_hrnetw18_small/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...n_hrnetw18_small/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:pp_humanseg_mobile
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/pp_humanseg_mobile/pp_humanseg_mobile_export_192x192
 --model_filename:model.pdmodel

diff --git a/..._tipc/configs/ocrnet_hrnetw18/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/..._tipc/configs/ocrnet_hrnetw18/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:ocrnet_hrnetw18
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/ocrnet_hrnetw18/ocrnet_hrnetw18_cityscapes_1024x512_160k
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/.../ocrnet_hrnetw18/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/.../ocrnet_hrnetw18/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:ocrnet_hrnetw18
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/ocrnet_hrnetw18/ocrnet_hrnetw18_cityscapes_1024x512_160k
 --model_filename:model.pdmodel

diff --git a/..._tipc/configs/ocrnet_hrnetw48/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/..._tipc/configs/ocrnet_hrnetw48/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:ocrnet_hrnetw48
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/ocrnet_hrnetw48/ocrnet_hrnetw48_cityscapes_1024x512_160k
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/.../ocrnet_hrnetw48/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/.../ocrnet_hrnetw48/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:ocrnet_hrnetw48
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/ocrnet_hrnetw48/ocrnet_hrnetw48_cityscapes_1024x512_160k
 --model_filename:model.pdmodel

diff --git a/...tipc/configs/pp_liteseg_stdc1/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/...tipc/configs/pp_liteseg_stdc1/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:pp_liteseg_stdc1
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/pp_liteseg_stdc1/pp_liteseg_infer_model
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/...pp_liteseg_stdc1/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...pp_liteseg_stdc1/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:pp_liteseg_stdc1
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/pp_liteseg_stdc1/pp_liteseg_infer_model
 --model_filename:model.pdmodel

diff --git a/...tipc/configs/pp_liteseg_stdc2/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/...tipc/configs/pp_liteseg_stdc2/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:pp_liteseg_stdc2
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/pp_liteseg_stdc2/pp_liteseg_stdc2_cityscapes_1024x512_scale1.0_160k
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/...pp_liteseg_stdc2/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...pp_liteseg_stdc2/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:pp_liteseg_stdc2
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/pp_liteseg_stdc2/pp_liteseg_stdc2_cityscapes_1024x512_scale1.0_160k
 --model_filename:model.pdmodel

diff --git a/..._tipc/configs/pphumanseg_lite/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/..._tipc/configs/pphumanseg_lite/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:pp_humanseg_lite
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/pp_humanseg_lite/pp_humanseg_lite_export_398x224
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:argmax_0.tmp_0
diff --git a/.../pphumanseg_lite/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/.../pphumanseg_lite/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:pp_humanseg_lite
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/pp_humanseg_lite/pp_humanseg_lite_export_398x224
 --model_filename:model.pdmodel

diff --git a/test_tipc/configs/ppmatting/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/test_tipc/configs/ppmatting/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:ppmatting
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/ppmatting/modnet-mobilenetv2
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:img
+--output_name:sigmoid_2.tmp_0
diff --git a/...onfigs/ppmatting/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...onfigs/ppmatting/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:ppmatting
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/ppmatting/modnet-mobilenetv2
 --model_filename:model.pdmodel

diff --git a/test_tipc/configs/stdc_stdc1/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt b/test_tipc/configs/stdc_stdc1/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt
@@ -0,0 +1,17 @@
+===========================serving_params===========================
+model_name:stdc_stdc1
+python:python3.7
+trans_model:-m paddle_serving_client.convert
+--dirname:./test_tipc/inferences/stdc_stdc1/stdc1seg_infer_model
+--model_filename:model.pdmodel
+--params_filename:model.pdiparams
+--serving_server:./test_tipc/serving_cpp/serving_server/
+--serving_client:./test_tipc/serving_cpp/serving_client/
+serving_dir:./test_tipc/serving_cpp
+--model:serving_server
+--op:GeneralSegOp
+--port:9997
+--gpu_id:"0"|null
+cpp_client:serving_client.py
+--input_name:x
+--output_name:save_infer_model/scale_0.tmp_1
diff --git a/...nfigs/stdc_stdc1/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt b/...nfigs/stdc_stdc1/serving_infer_python.txt → ...l_normal_serving_python_linux_gpu_cpu.txt
@@ -1,6 +1,6 @@
 ===========================serving_params===========================
 model_name:stdc_stdc1
-python:python3
+python:python3.7
 trans_model:-m paddle_serving_client.convert
 --dirname:./test_tipc/inferences/stdc_stdc1/stdc1seg_infer_model
 --model_filename:model.pdmodel

diff --git a/test_tipc/docs/test_serving_infer_cpp.md b/test_tipc/docs/test_serving_infer_cpp.md
@@ -0,0 +1,113 @@
+# Linux GPU/CPU CPP 服务化部署测试
+
+Linux GPU/CPU CPP 服务化部署测试的主程序为`test_serving_infer_cpp.sh`，可以测试基于CPP的模型服务化部署功能。
+
+
+## 1. 测试结论汇总
+
+- 推理相关：
+
+| 算法名称 | 模型名称 | device_CPU | device_GPU | batchsize |
+|  :----:   |  :----: |   :----:   |  :----:  |   :----:   |
+|  PP-LiteSeg   |  pp-liteSeg-stdc1 |  支持 | 支持 | 1 |
+|  PP-LiteSeg   |  pp-liteSeg-stdc2 |  支持 | 支持 | 1 |
+|  PP-Humanseg  |  pphumanseg_lite     |  支持 | 支持 | 1 |
+|  PP-Humanseg  |  pphumanseg_mobile   |  支持 | 支持 | 1 |
+|  PP-Humanseg  |  pphumanseg_server   |  支持 | 支持 | 1 |
+|  PP-Matting   |  pp_humanseg_matting |  支持 | 支持 | 1 |
+|  HRNet        |  fcn_hrnetw18     |  支持 | 支持 | 1 |
+|  OCRNet       |  ocrnet_hrnetw18  |  支持 | 支持 | 1 |
+|  OCRNet       |  ocrnet_hrnetw48  |  支持 | 支持 | 1 |
+|  STDCSeg      |  stdc_stdc1       |  支持 | 支持 | 1 |
+
+
+
+## 2. 测试流程
+
+### 2.1 准备环境
+
+* 配置docker
+
+拉取并进入Paddle Serving的GPU Docker (serving0.8.0-cuda10.2-cudnn7)。
+
+```
+nvidia-docker pull registry.baidubce.com/paddlepaddle/serving:0.8.0-cuda10.2-cudnn7-devel
+nvidia-docker run -p 9292:9292 --name test_serving_cpp -dit registry.baidubce.com/paddlepaddle/serving:0.8.0-cuda10.2-cudnn7-devel bash
+nvidia-docker exec -it test_serving_cpp bash
+```
+
+* 安装 PaddleServing
+
+安装PaddleServing最新版本的相关组件，包括serving_client、serving-app。
+PaddleServing的详细安装说明，参考[链接](https://github.com/PaddlePaddle/Serving/blob/v0.9.0/doc/Install_CN.md)。
+
+比如执行下面命令，在GPU CUDA 10.2环境下，安装0.8.3版本的PaddleServing组件。
+
+```
+pip3.7 install paddle-serving-client==0.8.3 -i https://pypi.tuna.tsinghua.edu.cn/simple
+pip3.7 install paddle-serving-app==0.8.3 -i https://pypi.tuna.tsinghua.edu.cn/simple
+```
+
+* 安装PaddlePaddle
+
+安装最新版本的PaddlePaddle，具体参考[文档](https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/install/pip/linux-pip.html)。
+
+比如执行下面命令，在GPU CUDA 10.2环境下，安装2.3.0版本PaddlePaddle。
+
+```
+pip3.7 install paddlepaddle-gpu==2.3.0 -i https://mirror.baidu.com/pypi/simple
+```
+
+* 准备PaddleSeg
+
+```
+git clone https://github.com/PaddlePaddle/PaddleSeg.git
+cd PaddleSeg
+git checkout -b develop origin/develop
+```
+
+### 2.2 编译安装paddle-serving-server
+
+在PaddleSeg根目录下，执行如下命令编译安装paddle-serving-server，耗时会较长。
+```
+sh test_tipc/serving_cpp/prepare_server.sh
+```
+
+在PaddleSeg根目录下，执行如下命令，设置环境变量。
+```
+export SERVING_BIN=${PWD}/Serving/build_server_gpu_opencv_seg/core/general-server/serving
+```
+
+### 2.2 功能测试
+
+测试方法如下所示，希望测试不同的模型文件，只需更换为对应的参数配置文件，即可完成对应模型的测试。
+
+```bash
+bash test_tipc/prepare.sh ${your_params_file} serving_infer
+bash test_tipc/test_serving_infer_cpp.sh ${your_params_file} serving_infer
+```
+
+以`PP-LiteSeg-Stdc1`的`Linux GPU/CPU CPP 服务化部署测试`为例，在PaddleSeg根目录下，执行如下命令。
+
+```bash
+bash test_tipc/prepare.sh test_tipc/configs/pp_liteseg_stdc1/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt serving_infer
+bash test_tipc/test_serving_infer_cpp.sh test_tipc/configs/pp_liteseg_stdc1/model_linux_gpu_normal_normal_serving_cpp_linux_gpu_cpu.txt serving_infer
+```
+
+log输出如下，表示测试成功。
+
+```
+Run successfully with command - pp_liteseg_stdc1 - python3.7 serving_client.py --input_name=x --output_name=argmax_0.tmp_0 > ../../log/pp_liteseg_stdc1/serving_infer/servering_infer_cpp_gpu_batchsize_1.log 2>&1 !
+
+...
+
+Run successfully with command - pp_liteseg_stdc1 - python3.7 serving_client.py --input_name=x --output_name=argmax_0.tmp_0 > ../../log/pp_liteseg_stdc1/serving_infer/servering_infer_cpp_cpu_batchsize_1.log 2>&1 !
+```
+
+GPU预测结果会自动保存在 `./log/pp_liteseg_stdc1/serving_infer/servering_infer_cpp_gpu_batchsize_1.log` ，CPU预测结果会保存在 `./log/pp_liteseg_stdc1/serving_infer/servering_infer_cpp_cpu_batchsize_1.log`，可以看到保存文件中有预测结果的均值（不同模型的value数值不同），表示测试成功。
+
+```
+result: {'res_mean_val': '[5.126434326171875]'}
+```
+
+如果运行失败，也会在终端中输出运行失败的日志信息以及对应的运行命令。可以基于该命令，分析运行失败的原因。