Update app.py
Browse files
app.py
CHANGED
@@ -1,11 +1,9 @@
|
|
1 |
from pprint import pprint
|
2 |
-
from
|
3 |
-
from haystack.nodes import BM25Retriever
|
4 |
-
from haystack.document_stores import InMemoryDocumentStore
|
5 |
-
from haystack.nodes import PromptTemplate, PromptNode
|
6 |
from PyPDF2 import PdfReader
|
7 |
import gradio as gr
|
8 |
import os
|
|
|
9 |
|
10 |
# Function to read PDF file content directly
|
11 |
def read_pdf(pdf_path):
|
@@ -19,38 +17,14 @@ def read_pdf(pdf_path):
|
|
19 |
def process_invoice(file, hf_token, questions):
|
20 |
# Read the PDF content directly
|
21 |
pdf_content = read_pdf(file.name)
|
22 |
-
document = Document(content=pdf_content)
|
23 |
-
docs = [document]
|
24 |
|
25 |
-
|
26 |
-
|
27 |
-
retriever = BM25Retriever(document_store, top_k=2)
|
28 |
-
|
29 |
-
qa_template = PromptTemplate(prompt=
|
30 |
-
""" Using exclusively the information contained in the context, answer only the question asked without adding
|
31 |
-
suggestions for possible questions, and respond exclusively in English. If the answer cannot be deduced from the
|
32 |
-
context, Don't add anything from the references if it is not asked explicitly. Do not repeat the same information twice
|
33 |
-
respond: "Not sure because not relevant to the context.
|
34 |
-
Context: {join(documents)};
|
35 |
-
Question: {query}
|
36 |
-
""")
|
37 |
-
|
38 |
-
prompt_node = PromptNode(
|
39 |
-
model_name_or_path='mistralai/Mixtral-8x7B-Instruct-v0.1',
|
40 |
-
api_key=hf_token,
|
41 |
-
default_prompt_template=qa_template,
|
42 |
-
max_length=500,
|
43 |
-
model_kwargs={"model_max_length": 5000}
|
44 |
-
)
|
45 |
-
|
46 |
-
rag_pipeline = Pipeline()
|
47 |
-
rag_pipeline.add_node(component=retriever, name="retriever", inputs=["Query"])
|
48 |
-
rag_pipeline.add_node(component=prompt_node, name="prompt_node", inputs=["retriever"])
|
49 |
|
50 |
answers = {}
|
51 |
for question in questions.split(','):
|
52 |
-
result =
|
53 |
-
answers[question] = result[
|
54 |
|
55 |
return answers
|
56 |
|
|
|
1 |
from pprint import pprint
|
2 |
+
from getpass import getpass
|
|
|
|
|
|
|
3 |
from PyPDF2 import PdfReader
|
4 |
import gradio as gr
|
5 |
import os
|
6 |
+
from transformers import pipeline
|
7 |
|
8 |
# Function to read PDF file content directly
|
9 |
def read_pdf(pdf_path):
|
|
|
17 |
def process_invoice(file, hf_token, questions):
|
18 |
# Read the PDF content directly
|
19 |
pdf_content = read_pdf(file.name)
|
|
|
|
|
20 |
|
21 |
+
# Initialize the Hugging Face pipeline
|
22 |
+
qa_pipeline = pipeline("question-answering", model="mistralai/Mixtral-8x7B-Instruct-v0.1", token=hf_token)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
answers = {}
|
25 |
for question in questions.split(','):
|
26 |
+
result = qa_pipeline(question=question.strip(), context=pdf_content)
|
27 |
+
answers[question] = result['answer']
|
28 |
|
29 |
return answers
|
30 |
|