Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
|
3 |
+
from langchain.document_loaders import OnlinePDFLoader
|
4 |
+
|
5 |
+
from langchain.text_splitter import CharacterTextSplitter
|
6 |
+
text_splitter = CharacterTextSplitter(chunk_size=350, chunk_overlap=0)
|
7 |
+
|
8 |
+
from langchain.llms import HuggingFaceHub
|
9 |
+
flan_ul2 = HuggingFaceHub(repo_id="google/flan-ul2", model_kwargs={"temperature":0.1, "max_new_tokens":300})
|
10 |
+
|
11 |
+
from langchain.embeddings import HuggingFaceHubEmbeddings
|
12 |
+
embeddings = HuggingFaceHubEmbeddings()
|
13 |
+
|
14 |
+
from langchain.vectorstores import Chroma
|
15 |
+
|
16 |
+
from langchain.chains import RetrievalQA
|
17 |
+
|
18 |
+
def infer(pdf_doc):
|
19 |
+
loader = OnlinePDFLoader(pdf_doc)
|
20 |
+
documents = loader.load()
|
21 |
+
texts = text_splitter.split_documents(documents)
|
22 |
+
db = Chroma.from_documents(texts, embeddings)
|
23 |
+
retriever = db.as_retriever()
|
24 |
+
qa = RetrievalQA.from_chain_type(llm=flan_ul2, chain_type="stuff", retriever=retriever, return_source_documents=True)
|
25 |
+
query = "What is the title of this paper?"
|
26 |
+
result = qa({"query": query})
|
27 |
+
|
28 |
+
return result
|
29 |
+
|
30 |
+
gr.Interface(fn=infer, inputs=[gr.Textbox(value="https://arxiv.org/pdf/2304.03757.pdf")], outputs=[gr.Textbox()]).launch()
|