Model_Manager/read_json.py at main · linsixi/Model_Manager · GitHub

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
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
import numpy as np
import matplotlib.pyplot as plt
import torch
import json
import requests
import argparse

# 自建使用函数
from judge_list import check_model_list  # 运行前检查函数
import MAS_Function  # MAS收敛函数
import model_judge as judge  # 判断模型类型并输出结果
import Model_api_new as api  # 调用大模型的函数
from download import download_file, model_json_url  # 下载模型函数
from DP_Function import differential_privacy_update

"""
##  符号解释
""" """         三个双引号代表其中的内容可以直接替换
#               井号表示注释

##  变量含义
in_type         判断输入的是图片或文字  文字：0  图片：1
value           串行中用来接收和返回模型的输入和输出
value_table     并行过程中用来接收所有模型的输出
file_path       读取的json文件的路径
model_type      模型的类型（是mindir还是onnx）
img_path        图片地址
weight_matrix   权重矩阵
url             文件在云端的地址

##  函数含义
read_json(file_path)            根据json文件来进行串行或并行操作
get_value(model_path)           分析模型后缀获取模型类型并调用模型
"""

AK = "F915WYG9INM7JCMKWYA8"
SK = "bWMR0xxcVBxOA6URk86efzREAOXLzoZvu6lkU00M"
ENDPOINT = "https://obs.cn-south-1.myhuaweicloud.com"

# 指定桶名称、文件键（OBS中文件的路径和文件名）、本地文件夹
bucket_name = "qg23onnx"
local_folder = "download"  # 你希望保存下载文件的本地文件夹


# 这里设置读取模型返回的值
def read_json(data):
    # 设置大模型对应的字典
    error = check_model_list(data)
    error_id = error["error_id"]
    error_message = error["message"]
    if error_id != 0:
        data.update({"error": error_message})
        return data
    switcher = {
        'get_qianfan_text': '1',
        'get_zidongtaichu': '2',
        'get_qianfan_graph': '3',
        'get_qianfan_read': '4'
    }

    # try:
    if not data["content"]:
        value = data["image"]
        in_type = 1  # 输入是图片
        image_url = value.split('https://qg23onnx.obs.cn-south-1.myhuaweicloud.com/')[-1]
        value = download_file(bucket_name, image_url, local_folder, ENDPOINT, AK, SK)  # value 改成下载的路径
        print(image_url)
        """这里要确定图片是给的地址，如有需要在这里添加读取图片的代码"""
    else:
        value = data["content"]  # 文字直接读取即可
        in_type = 0  # 输入是文字

    for layer_data in data["modelList"]:
        # 判断串并行条件
        if layer_data["parallel"] == 0:
            print("串行")

            for model in layer_data["models"]:  # 开始串行
                print(model["isAPI"])
                # 判断模型是本地还是调用大模型
                if model["isAPI"] == 0:  # 模型为本地
                    url = model["modelUrl"].split('https://qg23onnx.obs.cn-south-1.myhuaweicloud.com/')[-1]
                    model_path = download_file(bucket_name, url, local_folder, ENDPOINT, AK, SK)
                    model_map_url = model_json_url(url)
                    map_path = download_file(bucket_name, model_map_url, local_folder, ENDPOINT, AK, SK)
                    # 在函数内部判断模型是mindir还是onnx
                    value = judge.get_value(model["modelName"], model_path, value, in_type, map_path)

                else:  # 模型调用大模型
                    n = switcher.get(model["modelUrl"])
                    question_text = model["question"]
                    value = api.api_check(n, value, question_text)


        else:
            print("并行")  # 目前只支持文字输出

            value_tabel = []  # 用来接收所有输出
            weight_matrix = []  # 设置权重矩阵
            for model in layer_data["models"]:
                weight_matrix.append(model["weight"])

                # 判断模型是本地还是调用大模型
                if model["isAPI"] == 0:  # 模型为本地
                    url = model["modelUrl"].split('https://qg23onnx.obs.cn-south-1.myhuaweicloud.com/')[-1]
                    model_path = download_file(bucket_name, url, local_folder, ENDPOINT, AK, SK)
                    model_map_url = model_json_url(url)
                    map_path = download_file(bucket_name, model_map_url, local_folder, ENDPOINT, AK, SK)
                    value_tabel.append(
                        judge.get_value(model["modelName"], model_path, value, in_type, map_path))  # 判断是什么类型

                else:  # 模型调用大模型
                    n = switcher.get(model["modelUrl"])
                    question_text = model["question"]
                    value_tabel.append(api.api_check(n, value, question_text))
            weight_matrix = np.array(weight_matrix)
            noisy_matrix = differential_privacy_update(weight_matrix, iteration=1)
            # 归一化
            total_sum = np.sum(noisy_matrix)
            if total_sum != 0:
                noisy_matrix = noisy_matrix / total_sum
            else:
                noisy_matrix = np.zeros_like(weight_matrix)
            noisy_matrix.tolist()
            value = MAS_Function.change_value(value_tabel, noisy_matrix)

            print(value, "2222")
    """except Exception as e:
        print("错误类型：", type(e).__name__)
        print("错误信息：", str(e))"""
    print(value)
    data.update({"answer": str(value)})
    return data