Spaces:
Runtime error
Runtime error
Update main.py
Browse files
main.py
CHANGED
@@ -7,24 +7,45 @@ from typing import List
|
|
7 |
import pytesseract
|
8 |
import requests
|
9 |
from io import BytesIO
|
10 |
-
|
11 |
-
from llama_index.node_parser import SimpleNodeParser
|
12 |
-
llama-index
|
13 |
from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
|
14 |
|
|
|
|
|
15 |
|
16 |
|
17 |
|
18 |
|
19 |
description = """
|
20 |
## DocQA
|
21 |
-
|
22 |
This app shows how to do Document Question Answering
|
23 |
Check out the docs for the `/predict` endpoint below to try it out!
|
24 |
"""
|
25 |
|
26 |
app = FastAPI(docs_url="/", description=description)
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
# pipe = pipeline("document-question-answering", model="impira/layoutlm-document-qa")
|
29 |
|
30 |
|
@@ -64,9 +85,14 @@ def load_file(file_url: str, sentences: List[str]):
|
|
64 |
model_name = "deepset/roberta-base-squad2"
|
65 |
|
66 |
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
|
|
|
|
|
|
|
|
|
|
|
67 |
|
68 |
# Define the common context
|
69 |
-
context = all_text
|
70 |
|
71 |
# List of questions
|
72 |
questions = sentences
|
@@ -74,6 +100,7 @@ def load_file(file_url: str, sentences: List[str]):
|
|
74 |
qa_dict = {}
|
75 |
# Get answers for each question with the same context
|
76 |
for question in questions:
|
|
|
77 |
QA_input = {
|
78 |
'question': question,
|
79 |
'context': context
|
|
|
7 |
import pytesseract
|
8 |
import requests
|
9 |
from io import BytesIO
|
10 |
+
|
|
|
|
|
11 |
from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
|
12 |
|
13 |
+
from top2vec import Top2Vec
|
14 |
+
from llama_index.node_parser import SimpleNodeParser
|
15 |
|
16 |
|
17 |
|
18 |
|
19 |
description = """
|
20 |
## DocQA
|
|
|
21 |
This app shows how to do Document Question Answering
|
22 |
Check out the docs for the `/predict` endpoint below to try it out!
|
23 |
"""
|
24 |
|
25 |
app = FastAPI(docs_url="/", description=description)
|
26 |
|
27 |
+
def doc_chunk(data):
|
28 |
+
node_parser = SimpleNodeParser.from_defaults(chunk_size=256)
|
29 |
+
nodes = node_parser.get_nodes_from_documents(data)
|
30 |
+
return nodes
|
31 |
+
|
32 |
+
def create_train_data(nodes):
|
33 |
+
data = []
|
34 |
+
for i in range(len(nodes)):
|
35 |
+
#print(nodes[i].get_content())
|
36 |
+
data.append(nodes[i].get_content())
|
37 |
+
return data
|
38 |
+
|
39 |
+
def get_model(data):
|
40 |
+
model = Top2Vec(data, embedding_model='universal-sentence-encoder')
|
41 |
+
return model
|
42 |
+
|
43 |
+
def get_search_result(model, question):
|
44 |
+
documents, doc_scores, doc_ids = model.query_documents(question, 1)
|
45 |
+
|
46 |
+
return documents
|
47 |
+
|
48 |
+
|
49 |
# pipe = pipeline("document-question-answering", model="impira/layoutlm-document-qa")
|
50 |
|
51 |
|
|
|
85 |
model_name = "deepset/roberta-base-squad2"
|
86 |
|
87 |
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
|
88 |
+
##########################
|
89 |
+
nodes = doc_chunk(all_text)
|
90 |
+
data = create_train_data(nodes)
|
91 |
+
model = get_model(data)
|
92 |
+
#context = get_search_result(model, question)
|
93 |
|
94 |
# Define the common context
|
95 |
+
#context = all_text
|
96 |
|
97 |
# List of questions
|
98 |
questions = sentences
|
|
|
100 |
qa_dict = {}
|
101 |
# Get answers for each question with the same context
|
102 |
for question in questions:
|
103 |
+
context = get_search_result(model, question)
|
104 |
QA_input = {
|
105 |
'question': question,
|
106 |
'context': context
|