app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch
import random
import re

# Set manual seed for reproducibility
torch.manual_seed(42)

# Check for GPU availability
device = "cuda" if torch.cuda.is_available() else "cpu"

# Load the model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("humarin/chatgpt_paraphraser_on_T5_base")
model = AutoModelForSeq2SeqLM.from_pretrained(
    "humarin/chatgpt_paraphraser_on_T5_base"
).to(device)


# Function to paraphrase text
def humanize_text(text, temperature=0.7, max_length=512):
    input_ids = tokenizer(
        f"paraphrase: {text}",
        return_tensors="pt",
        padding=True,
        max_length=max_length,
        truncation=True,
    ).input_ids.to(device)

    # outputs = model.generate(
    #     input_ids,
    #     max_length=max_length,
    #     temperature=temperature,
    #     num_beams=1,
    #     num_beam_groups=1,
    #     num_return_sequences=1,
    #     repetition_penalty=2.0,
    #     diversity_penalty=0.5,
    #     no_repeat_ngram_size=2,
    # )

    outputs = model.generate(
        input_ids,
        max_length=max_length,
        do_sample=False,
        repetition_penalty=2.0,
        no_repeat_ngram_size=2,
    )

    paraphrased_texts = tokenizer.batch_decode(outputs, skip_special_tokens=True)
    return random.choice(paraphrased_texts)


# Function to split input into sentences
def split_into_sentences(text):
    return re.split(r"(?<=[.!?])\s+", text)


# Function to process multi-line text
def process_text(input_text):
    lines = input_text.split("\n")
    processed_lines = []

    for line in lines:
        if len(line) < 1:
            processed_lines.append(line)
        else:
            sentences = split_into_sentences(line)
            processed_sentences = [
                humanize_text(sentence, max_length=len(sentence))
                for sentence in sentences
            ]
            processed_lines.append(" ".join(processed_sentences))

    return "\n".join(processed_lines)


# Gradio Interface
iface = gr.Interface(
    fn=process_text,
    inputs=gr.Textbox(lines=5, placeholder="Enter text to humanize...", max_length=2000),
    outputs="text",
    title="AI Text Humanizer",
    description="Enter text, and the AI will rewrite it in a more human-like way.",
)

# Launch the Gradio app
if __name__ == "__main__":
    iface.launch()