app.py

import os
import streamlit as st
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import traceback

# Use Hugging Face Spaces' recommended persistent storage
CACHE_DIR = os.path.join(os.getcwd(), "model_cache")

def ensure_cache_dir():
    """
    Ensure the cache directory exists.
    
    Returns:
        str: Path to the cache directory
    """
    os.makedirs(CACHE_DIR, exist_ok=True)
    return CACHE_DIR

def load_model_and_tokenizer(model_name):
    """
    Load model and tokenizer with persistent caching.
    
    Args:
        model_name (str): Name of the model to load
    
    Returns:
        tuple: (model, tokenizer)
    """
    try:
        # Ensure cache directory exists
        cache_dir = ensure_cache_dir()
        
        # Construct full cache path for this model
        model_cache_path = os.path.join(cache_dir, model_name.replace('/', '_'))
        
        # Load tokenizer
        tokenizer = AutoTokenizer.from_pretrained(
            model_name, 
            cache_dir=model_cache_path
        )
        
        # Load model
        model = AutoModelForSeq2SeqLM.from_pretrained(
            model_name, 
            cache_dir=model_cache_path
        )
        
        return model, tokenizer
    except Exception as e:
        st.error(f"Error loading {model_name}: {str(e)}")
        st.error(traceback.format_exc())
        return None, None

def generate_summary(model, tokenizer, text, max_length=150):
    """
    Generate summary using a specific model and tokenizer.
    
    Args:
        model: Hugging Face model
        tokenizer: Hugging Face tokenizer
        text (str): Input text to summarize
        max_length (int): Maximum length of summary
    
    Returns:
        str: Generated summary
    """
    try:
        # Prepare input
        inputs = tokenizer(
            f"summarize: {text}", 
            max_length=512, 
            return_tensors="pt", 
            truncation=True
        )
        
        # Generate summary
        summary_ids = model.generate(
            inputs.input_ids, 
            num_beams=4, 
            max_length=max_length, 
            early_stopping=True
        )
        
        # Decode summary
        summary = tokenizer.decode(
            summary_ids[0], 
            skip_special_tokens=True
        )
        
        return summary
    except Exception as e:
        error_msg = f"Error in summarization: {str(e)}"
        st.error(error_msg)
        return error_msg

def main():
    st.title("Text Summarization with Pre-trained Models")
    
    # Display cache directory info (optional)
    st.info(f"Models will be cached in: {CACHE_DIR}")
    
    # Define models
    models_to_load = {
        'BART': 'facebook/bart-large-cnn',
        'T5': 't5-large',
        'Pegasus': 'google/pegasus-cnn_dailymail'
    }
    
    # Text input
    text_input = st.text_area("Enter text to summarize:")
    
    # Generate button
    if st.button("Generate Summary"):
        if not text_input:
            st.error("Please enter text to summarize.")
            return
        
        # Create columns for progressive display
        bart_col, t5_col, pegasus_col = st.columns(3)
        
        # Function to process each model
        def process_model(col, model_name, model_path):
            with col:
                with st.spinner(f'Generating {model_name} Summary...'):
                    progress = st.progress(0)
                    progress.progress(50)
                    
                    # Load model and tokenizer
                    model, tokenizer = load_model_and_tokenizer(model_path)
                    
                    if model and tokenizer:
                        # Generate summary
                        summary = generate_summary(model, tokenizer, text_input)
                        
                        progress.progress(100)
                        st.subheader(f"{model_name} Summary")
                        st.write(summary)
                    else:
                        st.error(f"Failed to load {model_name} model")
        
        # Process each model
        process_model(bart_col, 'BART', 'facebook/bart-large-cnn')
        process_model(t5_col, 'T5', 't5-large')
        process_model(pegasus_col, 'Pegasus', 'google/pegasus-cnn_dailymail')

if __name__ == "__main__":
    main()