from transformers import pipeline
import gradio as gr

pipe = pipeline(
    "text-generation",
    model="EleutherAI/MPT-7B-Instruct"  # or MPT-7B-Instruct for better results
)

def chat(user_input):
    output = pipe(
        user_input,
        max_length=100,
        do_sample=True,
        top_k=50,
        top_p=0.9,
        temperature=0.6
    )[0]['generated_text']
    return output

demo = gr.Interface(fn=chat, inputs="text", outputs="text")
demo.launch()