app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import spaces
from sentence_splitter import SentenceSplitter

device = "cuda"

tokenizer = AutoTokenizer.from_pretrained("NoaiGPT/777")
model = AutoModelForSeq2SeqLM.from_pretrained("NoaiGPT/777").to(device)

# Initialize the sentence splitter
splitter = SentenceSplitter(language='en')

@spaces.GPU
def generate_title(text):
    input_ids = tokenizer(f'paraphraser: {text}', return_tensors="pt", padding="longest", truncation=True, max_length=64).input_ids.to(device)
    outputs = model.generate(
        input_ids,
        num_beams=8,
        num_beam_groups=4,
        num_return_sequences=6,
        repetition_penalty=12.0,
        diversity_penalty=4.0,
        no_repeat_ngram_size=3,
        temperature=1.1,
        top_k=50,
        top_p=0.95,
        max_length=64
    )
    return tokenizer.batch_decode(outputs, skip_special_tokens=True)

def process_text(text):
    paragraphs = text.split('\n\n')
    results = []
    for paragraph in paragraphs:
        sentences = splitter.split(paragraph)
        paragraph_results = []
        for sentence in sentences:
            titles = generate_title(sentence)
            paragraph_results.append(f"Original: {sentence}\nParaphrases:\n" + "\n".join(titles))
        results.append("\n\n".join(paragraph_results))
    return "\n\n---\n\n".join(results)

def gradio_generate_title(text):
    return process_text(text)

iface = gr.Interface(
    fn=gradio_generate_title,
    inputs=gr.Textbox(lines=10, label="Input Text"),
    outputs=gr.Textbox(lines=20, label="Generated Paraphrases"),
    title="Diverse Paraphrase Generator",
    description="Generate multiple diverse paraphrases for each sentence in the input text using NoaiGPT/777 model."
)

iface.launch()