Spaces:
Sleeping
Sleeping
Azizul Hakim
commited on
app.py
Browse files
app.py
CHANGED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import torch
|
3 |
+
from transformers import PegasusForConditionalGeneration, PegasusTokenizer
|
4 |
+
|
5 |
+
# Caching the model and tokenizer loading to speed up the app
|
6 |
+
@st.cache_resource
|
7 |
+
def load_model_and_tokenizer():
|
8 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
9 |
+
|
10 |
+
model = PegasusForConditionalGeneration.from_pretrained('google/pegasus-large')
|
11 |
+
model.load_state_dict(torch.load('pegasus_bengali_epoch_2.pt')) # Load the saved model
|
12 |
+
model = model.to(device)
|
13 |
+
|
14 |
+
tokenizer = PegasusTokenizer('bengali_tokenizer.model', do_lower_case=False)
|
15 |
+
|
16 |
+
return model, tokenizer, device
|
17 |
+
|
18 |
+
# Summarization function
|
19 |
+
def summarize_bengali_text(model, tokenizer, device, text):
|
20 |
+
inputs = tokenizer(text, return_tensors="pt", max_length=512, truncation=True).to(device)
|
21 |
+
|
22 |
+
# Generate the summary
|
23 |
+
summary_ids = model.generate(
|
24 |
+
inputs['input_ids'],
|
25 |
+
max_length=70,
|
26 |
+
num_beams=5,
|
27 |
+
length_penalty=2.0,
|
28 |
+
early_stopping=True
|
29 |
+
)
|
30 |
+
|
31 |
+
summary = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
|
32 |
+
return summary
|
33 |
+
|
34 |
+
# Streamlit UI design
|
35 |
+
def main():
|
36 |
+
st.set_page_config(page_title="Bengali Text Summarizer", layout="centered", initial_sidebar_state="collapsed")
|
37 |
+
|
38 |
+
# Title and description
|
39 |
+
st.title("📜 Bengali Text Summarizer")
|
40 |
+
st.write("Enter a Bengali paragraph and get a concise summary instantly.")
|
41 |
+
|
42 |
+
# Sidebar info
|
43 |
+
with st.sidebar:
|
44 |
+
st.header("How to Use:")
|
45 |
+
st.write("1. Enter your Bengali text in the input box below.")
|
46 |
+
st.write("2. Click on **Summarize**.")
|
47 |
+
st.write("3. Get the summary of your input text.")
|
48 |
+
|
49 |
+
# User input
|
50 |
+
bengali_text = st.text_area("Enter Bengali Text:", height=200, placeholder="আপনার বাংলা টেক্সট এখানে লিখুন...")
|
51 |
+
|
52 |
+
# Load model and tokenizer
|
53 |
+
model, tokenizer, device = load_model_and_tokenizer()
|
54 |
+
|
55 |
+
# Summarization button
|
56 |
+
if st.button("Summarize"):
|
57 |
+
if bengali_text.strip():
|
58 |
+
# Summarize the text
|
59 |
+
with st.spinner("Summarizing..."):
|
60 |
+
summary = summarize_bengali_text(model, tokenizer, device, bengali_text)
|
61 |
+
|
62 |
+
# Display the result
|
63 |
+
st.subheader("Summary:")
|
64 |
+
st.write(summary)
|
65 |
+
else:
|
66 |
+
st.warning("Please enter some Bengali text to summarize!")
|
67 |
+
if __name__ == "__main__":
|
68 |
+
main()
|