PythonAssistant/RAG_ai/Copy.py at 7a2d535de3ae7035b6e293cecf31c372ded5fdfd · LLLLLLLLazy/PythonAssistant · 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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
import jieba
import numpy as np
from sentence_transformers import SentenceTransformer
import json
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.preprocessing import normalize
from xinference.client import RESTfulClient
from rank_bm25 import BM25Okapi
import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import ReadPDF

def AskAI(user_content):
    client = RESTfulClient("http://127.0.0.1:9997")
    model = client.get_model("chatglm3")
    messages = [{"role": "system", "content": '''你是一个汽车专家，帮我结合给定的资料，回答一个问题。如果问题无法从资料中获得，请输出结合给定的资料，无法回答问题。'''},
    {"role": "user", "content": user_content}]
    temp = model.chat(
        messages,
        generate_config={
            "max_tokens": 512,
            "temperature": 0.7
        }
    )
    return temp

pdf_content = ReadPDF.Readpdf()
questions = ReadPDF.Loadquestion()

# 对提问和PDF内容进行分词
# question_words = [' '.join(jieba.lcut(x['question'])) for x in questions]
# pdf_content_words = [' '.join(jieba.lcut(x['content'])) for x in pdf_content]
#
# tfidf = TfidfVectorizer()
# tfidf.fit(question_words + pdf_content_words)
#
# # 提取TFIDF
# question_feat = tfidf.transform(question_words)
# pdf_content_feat = tfidf.transform(pdf_content_words)
#
# # 进行归一化
# question_feat = normalize(question_feat)
# pdf_content_feat = normalize(pdf_content_feat)
#
# # 检索进行排序
# for query_idx, feat in enumerate(question_feat):
#     score = feat @ pdf_content_feat.T
#     score = score.toarray()[0]
#     max_score_page_idx = score.argsort()[-1] + 1
#     questions[query_idx]['reference'] = 'page_' + str(max_score_page_idx)

# #向量索引
# tokenizer = AutoTokenizer.from_pretrained('./Rerank/bge-reranker-base/')
# rerank_model = AutoModelForSequenceClassification.from_pretrained('./Rerank/bge-reranker-base/')
# rerank_model.cuda()

# pdf_content_words = [jieba.lcut(x['content']) for x in pdf_content]
# bm25 = BM25Okapi(pdf_content_words)
# count = 0
# for query_idx in range(len(questions)):
#     # 首先进行BM25检索
#     doc_scores = bm25.get_scores(jieba.lcut(questions[query_idx]["question"]))
#     max_score_page_idxs = doc_scores.argsort()[-3:]

#     # top3进行重排序
#     pairs = []
#     for idx in max_score_page_idxs:
#         pairs.append([questions[query_idx]["question"], pdf_content[idx]['content']])

#     inputs = tokenizer(pairs, padding=True, truncation=True, return_tensors='pt', max_length=512)
#     with torch.no_grad():
#         inputs = {key: inputs[key].cuda() for key in inputs.keys()}
#         scores = rerank_model(**inputs, return_dict=True).logits.view(-1, ).float()

#     max_score_page_idx = max_score_page_idxs[scores.cpu().numpy().argmax()]
#     questions[query_idx]['reference'] = 'page_' + str(max_score_page_idx + 1)
count = 0
model = SentenceTransformer('Embedding\\bge-small-zh-v1.5\\')
question_sentences = [x['question'] for x in questions]
pdf_content_sentences = [x['content'] for x in pdf_content]

question_embeddings = model.encode(question_sentences, normalize_embeddings=True)
pdf_embeddings = model.encode(pdf_content_sentences, normalize_embeddings=True)

for query_idx, feat in enumerate(question_embeddings):
    score = feat @ pdf_embeddings.T
    max_score_page_idx = score.argsort()[-1] + 1
    questions[query_idx]['reference'] = 'page_' + str(max_score_page_idx)
    prompt = '''你是一个汽车专家，帮我结合给定的资料，回答一个问题。如果问题无法从资料中获得，请输出结合给定的资料，无法回答问题。
    资料：{0}

    问题：{1}

    参考资料页码：{2}
        '''.format(
        pdf_content[max_score_page_idx]['content'],
        questions[query_idx]["question"],
        pdf_content[max_score_page_idx]['page']
    )
    print(prompt)
    answer = AskAI(prompt)['choices'][0]['message']['content']
    print(answer)
    questions[query_idx]['answer'] = answer

    count+=1
    if(count > 6):
        break


with open('testfiles\\submit.json', 'w', encoding='utf8') as up:
    json.dump(questions, up, ensure_ascii=False, indent=4)


# #BM25的方法
# pdf_content_words = [jieba.lcut(x['content']) for x in pdf_content]
# bm25 = BM25Okapi(pdf_content_words)
#
# for query_idx in range(len(questions)):
#     doc_scores = bm25.get_scores(jieba.lcut(questions[query_idx]["question"]))
#     max_score_page_idx = doc_scores.argsort()[-1] + 1
#     questions[query_idx]['reference'] = 'page_' + str(max_score_page_idx)

# with open('submit.json', 'w', encoding='utf8') as up:
#     json.dump(questions, up, ensure_ascii=False, indent=4)


# #向量索引
# model = SentenceTransformer('Embedding\\bge-small-zh-v1.5\\')
# question_sentences = [x['question'] for x in questions]
# pdf_content_sentences = [x['content'] for x in pdf_content]

# question_embeddings = model.encode(question_sentences, normalize_embeddings=True)
# pdf_embeddings = model.encode(pdf_content_sentences, normalize_embeddings=True)

# for query_idx, feat in enumerate(question_embeddings):
#     score = feat @ pdf_embeddings.T
#     max_score_page_idx = score.argsort()[-1] + 1
#     questions[query_idx]['reference'] = 'page_' + str(max_score_page_idx)

# with open('submit.json', 'w', encoding='utf8') as up:
#     json.dump(questions, up, ensure_ascii=False, indent=4)


# def ask_glm(content):
#     url = "https://open.bigmodel.cn/api/paas/v4/chat/completions"
#     headers = {
#       'Content-Type': 'application/json',
#       'Authorization': generate_token("填写key", 1000)
#     }
#
#     data = {
#         "model": "glm-3-turbo",
#         "messages": [{"role": "user", "content": content}]
#     }
#
#     response = requests.post(url, headers=headers, json=data)
#     return response.json()