app.py

import random
import gradio as gr
from transformers import pipeline


def random_response(message, history):
    return random.choice(["Yes", "No"])


generator = pipeline(model="michelebasilico/itaca-mistral-7b-v2-4bit")


def predict(message, history):
    outputs = generator(message, num_return_sequences=4,
                        return_full_text=False)

    return outputs


iface = gr.ChatInterface(predict)
iface.launch()