情感分析服务化部署之SimpleServing

[toc]

# 官方介绍文档

# 启动服务端

由于我机器paddlenlp没有配置环境变量，我们使用绝对路径运行/home/aistudio/paddlenlp2.6.0/bin/paddlenlp,默认启动脚本在安装包的bin/目录下

进入deploy目录

cd /home/aistudio/PaddleNLP/applications/sentiment_analysis/unified_sentiment_extraction/deploy

deploy讲解，该目录下有两个文件，一个server.py 一个client.py

server.py 命令中启动的服务就是这个文件

from paddlenlp import SimpleServer, Taskflow

# The schema changed to your defined schema
schema = [{"评价维度": ["观点词", "情感倾向[正向,负向,未提及]"]}]
# define taskflow to perform sentiment analysis
# 这里可以添加参数，指定我们自己模型的位置，task_path="/opt/model/nlp/model_best"
senta = Taskflow("sentiment_analysis", schema=schema, model="uie-senta-base")
# define your server
app = SimpleServer()
app.register_taskflow("taskflow/senta", senta)

1
2
3
4
5
6
7
8
9
10

client.py 客户端调用Demo

import json

import requests

url = "http://0.0.0.0:8189/taskflow/senta"
headers = {"Content-Type": "application/json"}
texts = ["蛋糕味道不错，店家的服务也很热情"]
data = {
 "data": {
     "text": texts,
 }
}
r = requests.post(url=url, headers=headers, data=json.dumps(data))
datas = json.loads(r.text)
print(datas)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

# 启动命令

/home/aistudio/paddlenlp2.6.0/bin/paddlenlp server server:app --workers 1 --host 0.0.0.0 --port 8189

我们也可以不进入deploy目录运行，通过--app_dir目录指定目录即可

/home/aistudio/paddlenlp2.6.0/bin/paddlenlp server server:app --workers 1 --host 0.0.0.0 --port 8189 --app_dir /home/aistudio/PaddleNLP/applications/sentiment_analysis/unified_sentiment_extraction/deploy

启动成功内如如下

$ /home/aistudio/paddlenlp2.6.0/bin/paddlenlp server server:app --workers 1 --host 0.0.0.0 --port 8189 --app_dir /home/aistudio/PaddleNLP/applications/sentiment_analysis/unified_sentiment_extraction/deploy
[2023-09-13 15:45:45,991] [    INFO] - starting to PaddleNLP SimpleServer...
[2023-09-13 15:45:45,991] [    INFO] - The PaddleNLP SimpleServer is starting, backend component uvicorn arguments as follows:
[2023-09-13 15:45:45,991] [    INFO] -    the starting argument [host]=0.0.0.0
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [port]=8189
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [log_level]=None
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [workers]=1
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [limit_concurrency]=None
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [limit_max_requests]=None
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [timeout_keep_alive]=15
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [app_dir]=/home/aistudio/PaddleNLP/applications/sentiment_analysis/unified_sentiment_extraction/deploy
[2023-09-13 15:45:45,992] [    INFO] -    the starting argument [reload]=False
[2023-09-13 15:45:52,612] [    INFO] - We are using <class 'paddlenlp.transformers.ernie.tokenizer.ErnieTokenizer'> to load '/home/aistudio/.paddlenlp/taskflow/sentiment_analysis/uie-senta-base'.
[2023-09-13 15:45:52,642] [    INFO] -    Taskflow  request [path]=/taskflow/senta is genereated.
INFO:     Started server process [1964]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8189 (Press CTRL+C to quit)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

# 参数说明

--host: 启动服务化的IP地址，通常可以设置成 0.0.0.0
--port：启动服务化的网络端口
--workers: 接收服务化的进程数，默认为1
--log_level：服务化输出日志的级别，默认为 info 级别
--limit_concurrency：服务化能接受的并发数目，默认为None, 没有限制
--timeout_keep_alive：保持服务化连接的时间，默认为15s
--app_dir：服务化本地的路径，默认为服务化启动的位置
--reload: 当 app_dir的服务化相关配置和代码发生变化时，是否重启server，默认为False

1
2
3
4
5
6
7
8

# 客户端调用

# 多条数据调用

在 texts 传入多条数据即可，返回的数据也是一个数组，对应上数组位置即可

import json

import requests

url = "http://0.0.0.0:8189/taskflow/senta"
headers = {"Content-Type": "application/json"}
texts = ["蛋糕味道不错，店家的服务也很热情","服务员不热情，服务态度不好，下次不来了"]
data = {
    "data": {
        "text": texts,
    }
}
r = requests.post(url=url, headers=headers, data=json.dumps(data))
datas = json.loads(r.text)
print(datas)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

结果

{
    "result": [
        {
            "评价维度": [
                {
                    "end": 4,
                    "probability": 0.8804405736499632,
                    "relations": {
                        "情感倾向[正向,负向,未提及]": [
                            {
                                "probability": 0.9998830584822258,
                                "text": "正向"
                            }
                        ],
                        "观点词": [
                            {
                                "end": 6,
                                "probability": 0.9943114289210939,
                                "start": 4,
                                "text": "不错"
                            }
                        ]
                    },
                    "start": 2,
                    "text": "味道"
                },
                {
                    "end": 12,
                    "probability": 0.995827464498646,
                    "relations": {
                        "情感倾向[正向,负向,未提及]": [
                            {
                                "probability": 0.9999358663076237,
                                "text": "正向"
                            }
                        ],
                        "观点词": [
                            {
                                "end": 16,
                                "probability": 0.9947389572457865,
                                "start": 14,
                                "text": "热情"
                            }
                        ]
                    },
                    "start": 10,
                    "text": "服务"
                }
            ]
        },
        {
            "评价维度": [
                {
                    "end": 2,
                    "probability": 0.9997505104257698,
                    "relations": {
                        "情感倾向[正向,负向,未提及]": [
                            {
                                "probability": 0.9999443299277857,
                                "text": "正向"
                            }
                        ],
                        "观点词": [
                            {
                                "end": 4,
                                "probability": 0.999357408362016,
                                "start": 3,
                                "text": "好"
                            }
                        ]
                    },
                    "start": 0,
                    "text": "蛋糕"
                },
                {
                    "end": 12,
                    "probability": 0.9861312855031521,
                    "relations": {
                        "情感倾向[正向,负向,未提及]": [
                            {
                                "probability": 0.9996660628408236,
                                "text": "负向"
                            }
                        ],
                        "观点词": [
                            {
                                "end": 14,
                                "probability": 0.3098063392794046,
                                "start": 12,
                                "text": "不好"
                            }
                        ]
                    },
                    "start": 10,
                    "text": "服务"
                }
            ]
        }
    ]
}

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
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101

# 自定义`schema`调用

默认我们再server.py代码中指定了schema,同时在Client调用时也可以支持定义 schema 传入到client请求中，可以快速切换 schema

这

import json

import requests

url = "http://0.0.0.0:8189/taskflow/senta"
headers = {"Content-Type": "application/json"}
texts = ["蛋糕味道不错，店家的服务也很热情","服务员不热情，服务态度不好，下次不来了"]
data = {
    "data": {
        "text": texts,
    },
    "parameters": {
        "schema": ["观点词"] # 自定义schema
    }
}
r = requests.post(url=url, headers=headers, data=json.dumps(data))
datas = json.loads(r.text)
print(datas)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

返回结果

{
    "result": [
        {
            "观点词": [
                {
                    "end": 16,
                    "probability": 0.9935872307286964,
                    "start": 14,
                    "text": "热情"
                },
                {
                    "end": 6,
                    "probability": 0.9989313489162441,
                    "start": 4,
                    "text": "不错"
                }
            ]
        },
        {
            "观点词": [
                {
                    "end": 13,
                    "probability": 0.9952653178475401,
                    "start": 11,
                    "text": "不好"
                },
                {
                    "end": 6,
                    "probability": 0.5518675358377116,
                    "start": 3,
                    "text": "不热情"
                }
            ]
        }
    ]
}

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

上次更新: 2023/12/13, 14:39:09

← 情感分析环境搭建之GPU版本情感分析服务化部署之构建Docker服务→