app.py

import gradio as gr
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

# Check if GPU is available, otherwise use CPU
device = "cuda" if torch.cuda.is_available() else "cpu"

# Load pre-trained GPT-2 model and tokenizer
model_name = "gpt2-large"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name).to(device)


def generate_text(input_text, max_length=16, num_beams=5, do_sample=False, no_repeat_ngram_size=2):
    """
    Generate text based on the given input text.

    Parameters:
    - input_text (str): The input text to start generation from.
    - max_length (int): Maximum length of the generated text.
    - num_beams (int): Number of beams for beam search.
    - do_sample (bool): Whether to use sampling or not.
    - no_repeat_ngram_size (int): Size of the n-gram to avoid repetition.

    Returns:
    - generated_text (str): The generated text.
    """
    # Encode the input text and move it to the appropriate device
    input_ids = tokenizer(input_text, return_tensors='pt')['input_ids'].to(device)
    # Generate text using the model
    output = model.generate(input_ids, max_length=max_length, num_beams=num_beams,
                            do_sample=do_sample, no_repeat_ngram_size=no_repeat_ngram_size)
    # Decode the generated output
    generated_text = tokenizer.decode(output[0])
    return generated_text


def generate_text_with_nucleus_search(input_text, max_length=16, do_sample=True, top_p=0.9):
    """
    Generate text with nucleus sampling based on the given input text.

    Parameters:
    - input_text (str): The input text to start generation from.
    - max_length (int): Maximum length of the generated text.
    - do_sample (bool): Whether to use sampling or not.
    - top_p (float): Nucleus sampling parameter.

    Returns:
    - generated_text (str): The generated text.
    """
    # Encode the input text and move it to the appropriate device
    input_ids = tokenizer(input_text, return_tensors='pt')['input_ids'].to(device)
    # Generate text using nucleus sampling
    output = model.generate(input_ids, max_length=max_length, do_sample=do_sample, top_p=top_p)
    # Decode the generated output
    generated_text = tokenizer.decode(output[0])
    return generated_text


# Create Gradio interface
input_textbox = gr.Textbox(lines=7, label="Input Text", placeholder="Enter your text here...")
output_textbox = gr.Textbox(label="Generated Text", placeholder="Generated text will appear here...")

gr.Interface(
    [generate_text, generate_text_with_nucleus_search],
    inputs=input_textbox,
    outputs=output_textbox,
    title="Text Generation with GPT-2",
    description="Enter some text and generate new text using GPT-2 model.",
    allow_flagging=False
).launch(share=True)