Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import pdfplumber
|
3 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
4 |
+
|
5 |
+
# Load the model
|
6 |
+
@st.cache_resource
|
7 |
+
def load_model():
|
8 |
+
tokenizer = AutoTokenizer.from_pretrained("OFA-Sys/mineru-base")
|
9 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("OFA-Sys/mineru-base")
|
10 |
+
return tokenizer, model
|
11 |
+
|
12 |
+
tokenizer, model = load_model()
|
13 |
+
|
14 |
+
# UI
|
15 |
+
st.title("📄 MinerU: Ask Questions from PDF")
|
16 |
+
uploaded_file = st.file_uploader("Upload a PDF", type="pdf")
|
17 |
+
question = st.text_input("Enter your question:")
|
18 |
+
|
19 |
+
if uploaded_file and question:
|
20 |
+
with pdfplumber.open(uploaded_file) as pdf:
|
21 |
+
text = ''
|
22 |
+
for page in pdf.pages:
|
23 |
+
text += page.extract_text()
|
24 |
+
|
25 |
+
# Prepare input for MinerU (usually expects a prompt)
|
26 |
+
input_text = f"question: {question} context: {text[:3000]}" # MinerU has token limit
|
27 |
+
|
28 |
+
inputs = tokenizer(input_text, return_tensors="pt", truncation=True)
|
29 |
+
outputs = model.generate(**inputs, max_new_tokens=128)
|
30 |
+
answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
31 |
+
|
32 |
+
st.markdown(f"### Answer:\n{answer}")
|