-
Notifications
You must be signed in to change notification settings - Fork 26
/
solr_retriever_reranker.py
193 lines (165 loc) · 7.41 KB
/
solr_retriever_reranker.py
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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
import json
import re
import requests
import os
from bs4 import BeautifulSoup
import pandas as pd
import numpy as np
from IPython.display import display, HTML
# Run ColBERT Reranker
from primeqa.components.reranker.colbert_reranker import ColBERTReranker
model_name_or_path = "DrDecr.dnn"
llmToken = os.getenv('LLM_TOKEN')
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'
# Retrieve documents
max_num_documents=10
def skip_unwanted_characters(document, keyword):
lines = document.split('\n')
desired_text = ""
last_occurrence = -1
for i, line in enumerate(lines):
if keyword in line:
last_occurrence = i
if last_occurrence != -1:
for line in lines[last_occurrence+1:]:
desired_text += line.strip() + "\n"
else:
desired_text = document
return desired_text.strip()
def pre_processingtext(text_data):
replaced = re.sub("\{{ .*?\}}", "", text_data)
replaced = re.sub("\{: .*?\}", "", replaced)
replaced = re.sub("\.*?", "", replaced)
replaced = re.sub("\(.*?\)|\[.*?\] |\{.*?\}", "", replaced)
replaced = re.sub("</?div[^>]*>", "", replaced)
replaced = re.sub("</?p[^>]*>", "", replaced)
replaced = re.sub("</?a[^>]*>", "", replaced)
replaced = re.sub("</?h*[^>]*>", "", replaced)
replaced = re.sub("</?em*[^>]*>", "", replaced)
replaced = re.sub("</?img*[^>]*>", "", replaced)
replaced = re.sub("&", "", replaced)
replaced = re.sub("</?href*>", "", replaced)
replaced = re.sub("\s+", " ", replaced)
replaced = replaced.replace("}","")
replaced = replaced.replace("##","")
replaced = replaced.replace("###","")
replaced = replaced.replace("#","")
replaced = replaced.replace("*","")
replaced = replaced.replace("<strong>","")
replaced = replaced.replace("</strong>","")
replaced = replaced.replace("<ul>","")
replaced = replaced.replace("</ul>","")
replaced = replaced.replace("<li>","")
replaced = replaced.replace("</li>","")
replaced = replaced.replace("<ol>","")
replaced = replaced.replace("</ol>","")
return replaced
# Retrieve documents
max_num_documents=10
def solr_reteriver(question):
question = question.replace("?","")
response = requests.get(f'http://150.239.171.68:8983/solr/superknowa/select?q='+question+'&q.op=AND&wt=json')
query_result = response.json()
# print("SOLR RESPONSE:", json.dumps(query_result, indent=2))
print(query_result['response']['numFound'], "documents found.")
total = query_result['response']['numFound']
results_list=[]
query_hits={}
if total > 0:
if total > 10:
total =10
for i in range(total):
string_unicode = query_result['response']['docs'][i]['content'][0]
doc = string_unicode.encode("ascii", "ignore")
string_decode = doc.decode()
keyword = "{: shortdesc} "
cleaned_text = skip_unwanted_characters(string_decode, keyword)
pattern = r'\{\s*:\s*[\w#-]+\s*\}|\{\s*:\s*\w+\s*\}|\n\s*\n'
cleaned_text = re.sub(pattern, '', cleaned_text)
cleaned_text = pre_processingtext(cleaned_text)
query_hits = {
"document": {
"rank": i,
"document_id": query_result['response']['docs'][i]['id'][0],
"text": cleaned_text[0:4000],
"url" :query_result['response']['docs'][i]['url'][0].replace(" ","")
},
}
results_list.append(query_hits)
results_to_display = [results_list['document'] for results_list in results_list]
df = pd.DataFrame.from_records(results_to_display, columns=['rank','document_id','text','url'])
# df['title'] = np.random.randint(1, 10, df.shape[0])
df.dropna(inplace=True)
print('======================================================================')
print(f'QUERY: {question}')
return results_list
def solr_reranker(question, max_reranked_documents = 10):
reranker = ColBERTReranker(model=model_name_or_path)
reranker.load()
results_list = solr_reteriver(question)
if len(results_list) >0:
reranked_results = reranker.predict(queries= [question], documents = [results_list], max_num_documents=max_reranked_documents)
print(reranked_results)
reranked_results_to_display = [result['document'] for result in reranked_results[0]]
df = pd.DataFrame.from_records(reranked_results_to_display, columns=['rank','document_id','text','url'])
print('======================================================================')
print(f'QUERY: {question}')
display( HTML(df.to_html()) )
return df['text'][0] , df['url'][0]
else:
return "0 documents found" , "None"
def format_string(doc):
doc = doc.encode("ascii", "ignore")
string_decode = doc.decode()
cleantext = BeautifulSoup(string_decode, "lxml").text
perfecttext = " ".join(cleantext.split())
perfecttext = re.sub(' +', ' ', perfecttext).strip('"')
# perfecttext = perfecttext[0:4000]
return perfecttext
def process_llm_request(question):
wd_result,url = solr_reranker(question)
if '0 documents found' not in wd_result:
combined_input = "Answer the question based only on the context below. " + \
"Context: " + format_string(wd_result) + \
" Question: " + question
print("INPUT PROMPT: ", combined_input)
headers = {
'Content-Type': 'application/json',
'Authorization': llmToken,
}
json_data = {
'model_id': 'bigscience/bloom',
# 'inputs': [
# messageText,
# ],
'inputs': [combined_input],
# "inputs": ["Answer the question based only on the context below. \
# Context: IBM Cloud Pak for Data offers the IBM Watson Knowledge Catalog service, which provides a number of features to incorporate such policy security, and compliance features and to govern your data. A data steward or administrator can use the IBM Watson Knowledge Catalog to build a governance catalog consisting of terms policies, and rules that can help govern and secure the data. \
# Question: What is Watson Knowledge catalog?"],
'parameters': {
# "stream": "true",
'temperature': 0.5,
'max_new_tokens': 200,
},
}
## demo link of llm server will be replace with original link
response = requests.post('https://llm-api.res.ibm.com/v1/generate', headers=headers, json=json_data)
json_response = json.loads(response.content.decode("utf-8"))
result = json_response['results'][0]['generated_text'].split("Answer:")
if len(result) > 1:
print("LLM Output: ", result[1])
return result[1],url
else:
print("LLM Output: ", result[0])
return result[0],url
else:
return "0 documents found" , "None"
def main():
question = "what is ibm cloud pak for data"
# print("-------- Final answer ---------------")
answer ,url = process_llm_request(question)
print("FINAL ANSWER: ", answer)
print("URL: ", url)
return answer , url
if __name__ == "__main__":
main()