src/summarizer.py

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch
import logging

class TextSummarizer:
    def __init__(self, model_name="facebook/bart-large-cnn"):
        """
        Initialize summarization model directly without using pipeline
        
        Args:
            model_name (str): Hugging Face model for summarization
        """
        try:
            # Force CPU usage and disable GPU
            self.device = torch.device('cpu')
            
            # Initialize tokenizer and model separately
            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
            self.model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
            
            # Move model to CPU and eval mode
            self.model = self.model.to(self.device)
            self.model.eval()
            
            logging.info("Summarization model initialized successfully")
            
        except Exception as e:
            logging.error(f"Failed to load summarization model: {str(e)}")
            raise RuntimeError(f"Failed to load summarization model: {str(e)}")
    
    def generate_summary(self, text, max_length=400, min_length=100):
        """
        Generate summary for given text
        
        Args:
            text (str): Input text to summarize
            max_length (int): Maximum length of summary
            min_length (int): Minimum length of summary
            
        Returns:
            str: Generated summary
        """
        try:
            # Validate input text
            if not text or len(text.strip()) == 0:
                return "No text provided for summarization."
            
            # Ensure min_length is less than max_length
            min_length = min(min_length, max_length)
            
            # Process text in chunks due to length limitations
            max_chunk_length = 1024  # BART's max input length
            chunks = [text[i:i + max_chunk_length] for i in range(0, len(text), max_chunk_length)]
            summaries = []
            
            with torch.no_grad():  # Disable gradient calculation
                for chunk in chunks:
                    if chunk.strip():
                        # Tokenize
                        inputs = self.tokenizer(chunk, max_length=1024, truncation=True, 
                                             return_tensors="pt")
                        inputs = inputs.to(self.device)
                        
                        # Generate summary
                        summary_ids = self.model.generate(
                            inputs["input_ids"],
                            num_beams=4,
                            max_length=max_length // len(chunks),
                            min_length=min_length // len(chunks),
                            length_penalty=2.0,
                            early_stopping=True
                        )
                        
                        # Decode summary
                        summary = self.tokenizer.decode(summary_ids[0], 
                                                      skip_special_tokens=True)
                        summaries.append(summary)
            
            return " ".join(summaries)
            
        except Exception as e:
            logging.error(f"Error during summarization: {str(e)}")
            return f"Error during summarization: {str(e)}"