app.py

import streamlit as st
import time
import requests

API_URL = "https://api-inference.huggingface.co/models/tuner007/pegasus_summarizer"
headers = {"Authorization": "Bearer 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"}

def query(payload):
	response = requests.post(API_URL, headers=headers, json=payload)
	return response.json()

@st.cache_resource 
def load_topic_transfomers():
  from transformers import pipeline
  try:
      topic_classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli",device="cuda", compute_type="float16")
  except Exception as e:
      topic_classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
      print("Error: ", e)
  return topic_classifier

def suggest_topic(topic_classifier,text):

    # while len(text)> 1024:
    #     text = summarize(whole_text[:-10])

    possible_topics = ["Gadgets", 'Business','Finance', 'Health', 'Sports',  'Politics','Government','Science','Education', 'Travel', 'Tourism', 'Finance & Economics','Market','Technology','Scientific Discovery',
                      'Entertainment','Environment','News & Media', "Space,Universe & Cosmos", "Fashion", "Manufacturing and Constructions","Law & Crime","Motivation", "Development & Socialization",  "Archeology"]

    result = topic_classifier(text, possible_topics)

    return result['labels']

st.title("Topic Suggestion")

if 'topic_model' not in st.session_state:
    with st.spinner("Loading Model....."):
        st.session_state.topic_model =  load_topic_transfomers() 
        st.success("Model_loaded")
    st.session_state.model = True

whole_text = st.text_input("Enter the text Here: ")

try:
    if st.button('Suggest topic'):
        start= time.time()  
                	
        output = query({
        	"inputs": whole_text,
        })
        st.subheader('Original Text: ')
        st.write(whole_text)
        
        st.subheader('\nSummarized Text:')
        st.write(output[0]["summary_text"])
        
        with st.spinner("Scanning content to suggest topics"):
            topic_classifier = st.session_state.topic_model 
            predicted_topic = suggest_topic(topic_classifier,whole_text)
        clk = time.time()-start
        if clk < 60:
            st.write(f'Generated in {(clk)} secs')
        else:
            st.write(f'Generated in {(clk)/60} minutes')
            
        st.subheader('Top 10 Topics related to the content')
        for i in predicted_topic[:10]:
            st.write(i)
except Exception as e:
    print("Error", e)