refine code & docs

Author: HydrogenSulfate

deploy/python_serving/pipeline_rpc_client.py

@@ -18,65 +18,8 @@
     from paddle_serving_server.pipeline import PipelineClient
 
 import argparse
-import base64
-import os
-import os.path as osp
 
-import cv2
-import numpy as np
-
-
-def numpy_to_base64(array: np.ndarray) -> str:
-    """numpy_to_base64
-
-    Args:
-        array (np.ndarray): input ndarray.
-
-    Returns:
-        bytes object: encoded str.
-    """
-    return base64.b64encode(array).decode('utf8')
-
-
-def video_to_numpy(file_path: str) -> np.ndarray:
-    """decode video with cv2 and return stacked frames
-       as numpy.
-
-    Args:
-        file_path (str): video file path.
-
-    Returns:
-        np.ndarray: [T,H,W,C] in uint8.
-    """
-    cap = cv2.VideoCapture(file_path)
-    videolen = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
-    decoded_frames = []
-    for i in range(videolen):
-        ret, frame = cap.read()
-        # maybe first frame is empty
-        if ret is False:
-            continue
-        img = frame[:, :, ::-1]
-        decoded_frames.append(img)
-    decoded_frames = np.stack(decoded_frames, axis=0)
-    return decoded_frames
-
-
-def parse_file_paths(input_path: str) -> list:
-    assert osp.exists(input_path), \
-        f"{input_path} did not exists!"
-    if osp.isfile(input_path):
-        files = [
-            input_path,
-        ]
-    else:
-        files = os.listdir(input_path)
-        files = [
-            file for file in files
-            if (file.endswith(".avi") or file.endswith(".mp4"))
-        ]
-        files = [osp.join(input_path, file) for file in files]
-    return files
+from utils import numpy_to_base64, parse_file_paths, video_to_numpy
 
 
 def parse_args():

deploy/python_serving/readme.md

@@ -123,18 +123,19 @@ fetch_var {
 
 ```
 ### 服务部署和请求
-paddleserving 目录包含了启动 pipeline 服务、C++ serving服务(TODO)和发送预测请求的代码，具体包括：
+`python_serving` 目录包含了启动 pipeline 服务、C++ serving服务(TODO)和发送预测请求的代码，具体包括：
 ```bash
 __init__.py
 configs/xxx.yaml            # 启动pipeline服务的配置文件
 pipeline_http_client.py     # http方式发送pipeline预测请求的python脚本
 pipeline_rpc_client.py      # rpc方式发送pipeline预测请求的python脚本
 recognition_web_service.py  # 启动pipeline服务端的python脚本
+utils.py                    # 储存预测过程中常用的函数，如parse_file_paths, numpy_to_base64, video_to_numpy
 ```
 #### Python Serving
 - 进入工作目录：
 ```bash
-cd deploy/paddleserving
+cd deploy/python_serving
 ```
 
 - 启动服务：
@@ -148,10 +149,10 @@ python3.7 recognition_web_service.py -n PPTSM -c configs/PP-TSM.yaml &>log.txt &
 - 发送请求：
 ```bash
 # 以http方式的发送预测请求并接受结果
-python3.7 pipeline_http_client.py
+python3.7 pipeline_http_client.py -i ../../data/example.avi
 
 # 以rpc方式的发送预测请求并接受结果
-python3.7 pipeline_rpc_client.py
+python3.7 pipeline_rpc_client.py -i ../../data/example.avi
 ```
 成功运行后，模型预测的结果会打印在 cmd 窗口中，结果如下：
 

deploy/python_serving/readme_en.md

@@ -123,22 +123,63 @@ fetch_var {
 
 ```
 ### Service deployment and requests
-The paddleserving directory contains the code for starting the pipeline service, C++ serving service (TODO) and sending prediction requests, including:
+The `python_serving` directory contains the code for starting the pipeline service, C++ serving service (TODO) and sending prediction requests, including:
 ```bash
 __init__.py
-configs/xxx.yaml # start the configuration file of the pipeline service
-pipeline_http_client.py # python script for sending pipeline prediction request via http
-pipeline_rpc_client.py # python script for sending pipeline prediction request in rpc mode
-recognition_web_service.py # python script that starts the pipeline server
+configs/xxx.yaml            # start the configuration file of the pipeline service
+pipeline_http_client.py     # python script for sending pipeline prediction request via http
+pipeline_rpc_client.py      # python script for sending pipeline prediction request in rpc mode
+recognition_web_service.py  # python script that starts the pipeline server
+utils.py                    # common functions used in inference, such as parse_file_paths, numpy_to_base64, video_to_numpy
 ```
 #### Python Serving
 - Go to the working directory:
 ```bash
-cd deploy/paddleserving
+cd deploy/python_serving
 ```
 
 - Start the service:
 ```bash
 # Start in the current command line window and stay in front
 python3.7 recognition_web_service.py -n PPTSM -c configs/PP-TSM.yaml
-# Start in the background, the logs printed during the process will be redirected and saved to lo
+# Start in the background, the logs printed during the process will be redirected and saved to log.txt
+python3.7 recognition_web_service.py -n PPTSM -c configs/PP-TSM.yaml &>log.txt &
+```
+
+- send request:
+```bash
+# Send a prediction request in http and receive the result
+python3.7 pipeline_http_client.py -i ../../data/example.avi
+
+# Send a prediction request in rpc and receive the result
+python3.7 pipeline_rpc_client.py -i ../../data/example.avi
+```
+After a successful run, the results of the model prediction will be printed in the cmd window, and the results are as follows:
+
+```bash
+# http method print result
+{'err_no': 0, 'err_msg': '', 'key': ['label', 'prob'], 'value': ["['archery']", '[0.9907388687133789]'], 'tensors ': []}
+
+# The result of printing in rpc mode
+PipelineClient::predict pack_data time:1645631086.764019
+PipelineClient::predict before time:1645631086.8485317
+key: "label"
+key: "prob"
+value: "[\'archery\']"
+value: "[0.9907388687133789]"
+```
+
+## FAQ
+**Q1**: No result is returned after the request is sent or an output decoding error is prompted
+
+**A1**: Do not set the proxy when starting the service and sending the request. You can close the proxy before starting the service and sending the request. The command to close the proxy is:
+```
+unset https_proxy
+unset http_proxy
+```
+
+**Q2**: There is no response after the server is started, and it has been stopped at `start proxy service`
+
+**A2**: It is likely that a problem was encountered during the startup process. You can view the detailed error message in the `./deploy/python_serving/PipelineServingLogs/pipeline.log` log file
+
+For more service deployment types, such as `RPC prediction service`, you can refer to Serving's [github official website](https://github.com/PaddlePaddle/Serving/tree/v0.7.0/examples)