app.py

import gradio as gr
from transformers import pipeline, AutoConfig
import logging

# Temporary workaround: Extend the Logger class to include warning_once
def warning_once(self, msg, *args, **kwargs):
    if msg not in self._warned:
        self.warning(msg, *args, **kwargs)
        self._warned.add(msg)

logging.Logger.warning_once = warning_once
logging.Logger._warned = set()

# Load the model configuration with trust_remote_code to execute local configuration
config = AutoConfig.from_pretrained('tiiuae/falcon-40b-instruct', trust_remote_code=True)

# Load the model once when the script starts using the loaded config
generator = pipeline('text-generation', model='tiiuae/falcon-40b-instruct', config=config)

def generate_text(prompt):
    # Use the preloaded model
    return generator(prompt, max_length=100)[0]['generated_text']

def main():
    with gr.Blocks() as demo:
        gr.Markdown("## Text Generation Model")
        gr.Markdown("This model generates text based on the input prompt. Powered by Hugging Face transformers.")
        prompt = gr.Textbox(lines=2, placeholder="Type your prompt here...")
        output = gr.Text(label="Generated Text")

        prompt.change(fn=generate_text, inputs=prompt, outputs=output)

    demo.launch()

if __name__ == "__main__":
    main()