app.py

import gradio as gr
from huggingface_hub.utils import HfHubHTTPError

def predict(message, history):
    try:
        # Load the model interface directly using the load method
        chat_interface = gr.Interface.load("models/meta-llama/Meta-Llama-3.1-8B")

        # Get the prediction using the correct method call
        response = chat_interface.predict(message)
        history.append((message, response))
        return "", history

    except HfHubHTTPError as e:
        if e.response.status_code == 504:
            return "Server overloaded. Please try again later.", history
        else:
            raise e

with gr.Blocks() as demo:
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.ClearButton([msg, chatbot])

    msg.submit(predict, [msg, chatbot], [msg, chatbot])