-
Notifications
You must be signed in to change notification settings - Fork 0
/
rag_model.py
201 lines (172 loc) · 7.3 KB
/
rag_model.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
194
195
196
197
198
import os
import requests
import fitz
from tqdm.auto import tqdm
from spacy.lang.en import English
import numpy as np
import pandas as pd
import re
from sentence_transformers import util, SentenceTransformer
import random
import torch
from time import perf_counter as timer
import textwrap
from dotenv import load_dotenv
from groq import Groq
import warnings
warnings.filterwarnings('ignore')
device = 'cuda' if torch.cuda.is_available() else 'cpu'
embeddings_save_path = 'text_and_embeddings.csv'
pdf_path = './human-nutrition-text.pdf'
# pdf_path = 'PG_Information_Brochure_Jul_2021.pdf'
# model_name = 'all-mpnet-base-v2'
model_name = 'all-MiniLM-L6-v2'
model_llm_name = 'llama3-8b-8192'
def print_wrapped(text, wrap_length=80):
wrapped_text = textwrap.fill(text, wrap_length)
print(wrapped_text)
def preprocess():
pdf_url = 'https://www.iitk.ac.in/doaa/DOAA/PG_Information_Brochure_Jul_2021.pdf' # enter url here
if not os.path.exists(pdf_path):
print(f'Downloading file . . .')
response = requests.get(pdf_url)
if response.status_code == 200:
with open(pdf_path, "wb") as f:
f.write(response.content)
print(f'File is downloaded as {pdf_path}')
else:
print(f'[ERROR] {response.status_code}')
else:
print(f'File {pdf_path} already exists')
def formatter(text):
clean = text.replace("\n", " ").strip()
return clean
def open_read(pdf_path):
doc = fitz.open(pdf_path)
pages = []
for page_no, page in tqdm(enumerate(doc)):
text = formatter(page.get_text())
pages.append({"page_no" : page_no-41 ,
"page_char_count" : len(text),
"page_word_count" : len(text.split(" ")),
"page_sentence_count" : len(text.split(". ")),
"page_token_count" : len(text)/4,
"text": text})
return pages
print(f'Processing PDF {pdf_path} . . .')
pages = open_read(pdf_path=pdf_path)
print(f'Sentencizing . . .')
nlp = English()
nlp.add_pipe("sentencizer")
for item in tqdm(pages):
item["sentences"] = list(nlp(item["text"]).sents)
item["sentences"] = [str(x) for x in item["sentences"]]
item["sentence_count_spacy"] = len(item["sentences"])
print(f'Finished sentencizing \n')
print(f'Chunking . . .')
chunk_size = 10
def split_list(input_list: list[str],
slice_size: int = chunk_size) -> list[list[str]]:
return [input_list[i : i+slice_size] for i in range(0, len(input_list), slice_size)]
for item in tqdm(pages):
item["sentence_chunks"] = split_list(item["sentences"], chunk_size)
item["num_chunks"] = len(item["sentence_chunks"])
pages_chunks = []
for item in tqdm(pages):
for chunk in item["sentence_chunks"]:
chunk_dict = {}
chunk_dict["page_number"] = item["page_no"]
joined_chunk = "".join(chunk).replace(" ", " ").strip()
joined_chunk = re.sub(r'\.([A-Z])', r'. \1', joined_chunk) # ".A" -> ". A"
chunk_dict["sentence_chunk"] = joined_chunk
chunk_dict["chunk_char_count"] = len(joined_chunk)
chunk_dict["chunk_word_count"] = len([word for word in joined_chunk.split(" ")])
chunk_dict["chunk_token_count"] = len(joined_chunk) / 4
pages_chunks.append(chunk_dict)
df = pd.DataFrame(pages_chunks)
min_token_length = 30
pages_chunks_mod = df[df["chunk_token_count"] >= min_token_length].to_dict(orient="records")
print(f'Finished chunking \n')
print(f'Beginning embedding . . .')
start_time = timer()
embed_model = SentenceTransformer(model_name_or_path=model_name, device=device)
for item in tqdm(pages_chunks_mod):
item["embedding"] = embed_model.encode(item["sentence_chunk"])
print(f'Saving embeddings to {embeddings_save_path}')
text_and_embeddings = pd.DataFrame(pages_chunks_mod)
text_and_embeddings.to_csv(embeddings_save_path, index=False)
end_time = timer()
print(f'Finished embedding \n')
print(f'Time taken to embed : {end_time-start_time:.4f}')
def retrieve(query, embeddings, model=model_name, num_returns=5, print_time=False, print_results=False, data=None):
query_embedding = model.encode(query, convert_to_tensor=True).to(device=device)
start_time = timer()
dot_scores = util.dot_score(query_embedding, embeddings)[0]
end_time = timer()
scores, indices = torch.topk(dot_scores, k=num_returns)
if print_time:
print(f'Time to retrieve : {end_time-start_time:.6f}')
if print_results:
assert data
print(f'Query: {query}')
for score, idx in zip(scores, indices):
print(f'Score : {score}')
print(f'Text: \n')
print_wrapped(data[idx]["sentence_chunk"])
print(f'Page number: {data[idx]["page_number"]}')
print('\n')
return scores, indices
def prompt_formatter(query:str, context_items:list[dict]) -> str:
context = '- ' + "\n- ".join([item["sentence_chunk"] for item in context_items])
prompt = f"""Based on the following context items answer the query.
Give yourself room to think by extracting relevant passages from the context before answering the query.
Don't return the thinking, return only the answer.
Make sure your answers are as explanatory as possible.
Context_items: {context}
Query : {query}
Return your answers in the below format:
Query : <the query here>
Answer : <Your answer>
"""
return prompt
def ask(query, print_sources:bool=False):
start_time = timer()
text_and_embeddings = pd.read_csv(embeddings_save_path)
text_and_embeddings["embedding"] = text_and_embeddings["embedding"].apply(lambda x : np.fromstring(x.strip("[]"), sep=" "))
embeddings = torch.tensor(np.stack(text_and_embeddings["embedding"].to_list(), axis=0, dtype=np.float32)).to(device)
pages_chunks_mod = text_and_embeddings.to_dict(orient="records")
print(f'Embedding dimension : {embeddings.shape[1]}')
embed_model = SentenceTransformer(model_name_or_path=model_name, device=device)
load_dotenv()
client = Groq(
api_key=os.getenv("GROQ_API_KEY"),
)
_, indices = retrieve(query, embeddings, embed_model, print_time=True, data=pages_chunks_mod)
print(f'Retrieved relevant sources \n')
context_items = [pages_chunks_mod[i] for i in indices]
prompt = prompt_formatter(query, context_items=context_items)
chat_completion = client.chat.completions.create(
messages=[
{
"role": "user",
"content": prompt,
}
],
model=model_llm_name,
)
rag_answer = chat_completion.choices[0].message.content
end_time = timer()
print(f'Time taken to answer : {end_time-start_time:.4f}')
if print_sources:
print(f'Relevant sources are : \n')
for text in context_items:
print(f'- {text["sentence_chunk"]}')
print('\n\n')
return rag_answer.replace(prompt, '')
if __name__ == '__main__':
if not os.path.exists(embeddings_save_path):
preprocess()
print(f'Preprocessing done \n')
query = input('Enter your query now : ')
result = ask(query=query, print_sources=True)
print(result)