utils.py

import logging
from transformers import pipeline, T5Tokenizer, T5ForConditionalGeneration
from pdfminer.high_level import extract_text


logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
logger = logging.getLogger(__name__)

pipe = pipeline("text2text-generation", model="google-t5/t5-base", device="cpu")
pipe.model.config.pad_token_id = pipe.tokenizer.eos_token_id 

fine_tuned_model_path = "./fine_tuned_model"
fine_tuned_model = T5ForConditionalGeneration.from_pretrained(fine_tuned_model_path)
fine_tuned_tokenizer = T5Tokenizer.from_pretrained(fine_tuned_model_path)


def pdf_to_text(pdf_path):
    try:
        logger.info(f"Extracting text from PDF: {pdf_path}")
        return extract_text(pdf_path)
    except Exception as e:
        logger.error(f"Error while extracting text from PDF: {str(e)}")
        raise ValueError(f"PDF extraction error: {str(e)}")

def generate_lesson_from_transcript(doc_text):
    try:
        logger.info("Generating lesson from transcript using general model.")
        generated_text = pipe(doc_text, max_length=100, truncation=True)[0]['generated_text']
        output_path = "/tmp/generated_output.txt"
        
        with open(output_path, "w") as file:
            file.write(generated_text)
        
        logger.info(f"Lesson generation successful. Output saved at: {output_path}")
        return generated_text, output_path

    except Exception as e:
        logger.error(f"Error occurred during lesson generation: {str(e)}")
        return "An error occurred", None

def refine_with_fine_tuned_model(general_output):
    try:
        logger.info("Refining the output with fine-tuned model.")
        prompt = "Refine the following text for teaching purposes: " + general_output
        inputs = fine_tuned_tokenizer(prompt, return_tensors="pt", padding=True, truncation=True, max_length=512)
        output_ids = fine_tuned_model.generate(
            inputs["input_ids"],
            max_length=300,
            no_repeat_ngram_size=3,
            early_stopping=True
        )
        refined_text = fine_tuned_tokenizer.decode(output_ids[0], skip_special_tokens=True)
        return refined_text
    except Exception as e:
        logger.error(f"Error during refinement with fine-tuned model: {str(e)}")
        return "An error occurred during refinement."

def split_text_into_chunks(text, chunk_size=1000):
    words = text.split()
    chunks = []
    for i in range(0, len(words), chunk_size):
        chunk = ' '.join(words[i:i+chunk_size])
        chunks.append(chunk)
    return chunks

def generate_lesson_from_chunks(chunks):
    generated_texts = []
    for chunk in chunks:
        try:
            generated_text = pipe(chunk, max_length=500, truncation=True)[0]['generated_text']
            generated_texts.append(generated_text)
        except Exception as e:
            print(f"Error in chunk processing: {str(e)}")
            continue
    return ' '.join(generated_texts)

def process_large_text(text):
    chunks = split_text_into_chunks(text, chunk_size=1000)
    generated_text = generate_lesson_from_chunks(chunks)
    return generated_text