基于PaddleServing的 PaddlePaddle+PaddleOCR服务化部署(k8s)

发表于 更新于 阅读次数： 本文字数： 5.6k 阅读时长 ≈ 5 分钟

官方文档

https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/deploy/pdserving/README_CN.md

https://github.com/PaddlePaddle/Serving/blob/v0.8.3/doc/Latest_Packages_CN.md

PaddlePaddle是一个深度学习框架，提供了丰富的深度学习模型和工具，可以让用户快速构建、训练和部署深度学习模型。PaddlePaddle 支持多种编程语言，包括 Python、C++等，同时还提供了丰富的数据集和工具集，例如预训练模型、模型优化工具等。PaddlePaddle 可以用于各种深度学习应用，包括计算机视觉、自然语言处理、推荐系统等。

PaddleServing 则是一个用于部署深度学习模型的开源服务器，它可以将训练好的深度学习模型转换为可以被部署到生产环境中的格式，例如 HTTP API、Docker 容器等。PaddleServing 支持多种深度学习框架，包括 TensorFlow、PyTorch 等，同时还提供了丰富的功能，例如模型优化、量化、调参等。PaddleServing 可以帮助用户快速构建深度学习应用程序，并方便地进行部署和扩展。

部署

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
FROM registry.baidubce.com/paddlepaddle/paddle:2.4.2
COPY PaddleOCR-2.6.0.zip /home/
RUN unzip PaddleOCR-2.6.0.zip \
&& cd PaddleOCR-2.6.0 \
&& pip3 install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple \
&& cd deploy/pdserving \
&& wget https://paddle-serving.bj.bcebos.com/test-dev/whl/paddle_serving_server-0.8.3-py3-none-any.whl && pip3 install paddle_serving_server-0.8.3-py3-none-any.whl -i https://pypi.tuna.tsinghua.edu.cn/simple \
&& wget https://paddle-serving.bj.bcebos.com/test-dev/whl/paddle_serving_client-0.8.3-cp37-none-any.whl && pip3 install paddle_serving_client-0.8.3-cp37-none-any.whl -i https://pypi.tuna.tsinghua.edu.cn/simple \
&& wget https://paddle-serving.bj.bcebos.com/test-dev/whl/paddle_serving_app-0.8.3-py3-none-any.whl && pip3 install paddle_serving_app-0.8.3-py3-none-any.whl -i https://pypi.tuna.tsinghua.edu.cn/simple \
&& wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_infer.tar -O ch_PP-OCRv3_det_infer.tar && tar -xf ch_PP-OCRv3_det_infer.tar \
&& wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_rec_infer.tar -O ch_PP-OCRv3_rec_infer.tar &&  tar -xf ch_PP-OCRv3_rec_infer.tar \
&& python3 -m paddle_serving_client.convert --dirname ./ch_PP-OCRv3_det_infer/ \
                                         --model_filename inference.pdmodel          \
                                         --params_filename inference.pdiparams       \
                                         --serving_server ./ppocr_det_v3_serving/ \
                                         --serving_client ./ppocrv3_det_client/ \
&& python3 -m paddle_serving_client.convert --dirname ./ch_PP-OCRv3_rec_infer/ \
                                         --model_filename inference.pdmodel          \
                                         --params_filename inference.pdiparams       \
                                         --serving_server ./ppocr_rec_v3_serving/  \
                                         --serving_client ./ppocrv3_rec_client/
WORKDIR /home/PaddleOCR-2.6.0/deploy/pdserving
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
apiVersion: apps/v1
kind: Deployment
metadata:
  name: paddlepaddle-ocr
  namespace: paddlepaddle-ocr
spec:
  replicas: 1
  selector:
    matchLabels:
      app: paddlepaddle-ocr
  template:
    metadata:
      labels:
        app: paddlepaddle-ocr
    spec:
      containers:
        - name: paddlepaddle-ocr
          image: devops/paddlepaddle-ocr:v1
          command:
            - python3
          args:
            - 'web_service.py'
            - '/home/PaddleOCR-2.6.0/deploy/pdserving/--config=config.yml'
          ports:
            - name: 9998
              containerPort: 9998
              protocol: TCP
            - name: 18091
              containerPort: 18091
              protocol: TCP
          resources:
            limits:
              memory: 5Gi
            requests:
              memory: 5Gi
          terminationMessagePath: /dev/termination-log
          terminationMessagePolicy: File
          imagePullPolicy: IfNotPresent
      restartPolicy: Always
      terminationGracePeriodSeconds: 30
      dnsPolicy: ClusterFirst
      securityContext: {}
      schedulerName: default-scheduler
  strategy:
    type: Recreate
  revisionHistoryLimit: 10
  progressDeadlineSeconds: 600
---
apiVersion: v1
kind: Service
metadata:
  name: paddlepaddle-ocr
  namespace: paddlepaddle-ocr
spec:
  ports:
    - name: http
      protocol: TCP
      port: 9093
  selector:
    app: paddlepaddle-ocr
  type: NodePort
  sessionAffinity: None
  externalTrafficPolicy: Cluster

调用

部署后java后端调用时发现接口返回的数据为字符串，并且发现无法正常解析

官方示例中用了eval函数，但是java并没有

后面就用python中转了下

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
import json
from flask import Flask
from flask import request
import requests

app = Flask(__name__)


@app.route("/ocr/prediction", methods=["POST"])
def hello():
    img_b64 = request.json['img_b64']
    url = "http://paddlepaddle-ocr:9998/ocr/prediction"
    payload = json.dumps({
        "key": [
            "image"
        ],
        "value": [img_b64]
    })
    headers = {
        'Content-Type': 'application/json'
    }
    r = requests.request("POST", url, headers=headers, data=payload)
    result = r.json()
    ocr_result = result["value"][0]
    ret = []
    for item in eval(ocr_result):
        # print(item[0])
        ret.append(item[0][0])
    data = {
        "name": ret
    }
    print(ret)
    response = json.dumps(data)  # 将python的字典转换为json字符串
    return response, 200, {"Content-Type": "application/json"}

1
2
3
4
5
6
7
8
9
10
11
12
13
[uwsgi]
#uwsgi启动时，所使用的地址和端口（这个是http协议的）
http=0.0.0.0:80
#指向程序目录
chdir=/root
#python 启动程序文件
wsgi-file=app.py
#python 程序内用以启动的application 变量名
callable=app
#处理器数
processes=4
#线程数
threads=2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
apiVersion: apps/v1
kind: Deployment
metadata:
  name: paddlepaddle-ocr-api
  namespace: paddlepaddle-ocr
spec:
  replicas: 1
  selector:
    matchLabels:
      app: paddlepaddle-ocr-api
  template:
    metadata:
      creationTimestamp: null
      labels:
        app: paddlepaddle-ocr-api
    spec:
      containers:
        - name: paddlepaddle-ocr-api
          image: devops/paddlepaddle-ocr-api:v1
          command:
            - uwsgi
          args:
            - '--ini'
            - /root/api.ini
          ports:
            - name: http
              containerPort: 80
              protocol: TCP
          resources: {}
          terminationMessagePath: /dev/termination-log
          terminationMessagePolicy: File
          imagePullPolicy: Always

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
FROM devops/python:3.7.16
COPY app.py /root/app.py
COPY api.ini /root/api.ini
RUN pip install -i https://pypi.tuna.tsinghua.edu.cn/simple \
certifi==2022.12.7 \
charset-normalizer==3.1.0 \
click==8.1.3 \
colorama==0.4.6 \
Flask==2.2.5 \
idna==3.4 \
importlib-metadata==6.6.0 \
itsdangerous==2.1.2 \
Jinja2==3.1.2 \
MarkupSafe==2.1.2 \
requests==2.28.2 \
typing-extensions==4.5.0 \
urllib3==1.26.15 \
Werkzeug==2.2.3 \
zipp==3.15.0 \
uwsgi==2.0.21