app.py

import gradio as gr
from huggingface_hub import InferenceClient
import spaces
import os

client = InferenceClient("meta-llama/Llama-2-70b-chat-hf", token=f"Bearer {os.environ.get('API_KEY')}")
messages=[]
client.chat_completion(messages, max_tokens=1024)

@spaces.GPU()
def respond(prompt):
    response = client.chat_completion(
        model="meta-llama/Meta-Llama-3-70B-Instruct",
        messages=messages,
        max_tokens=500,
    )
    return response.content

gr.ChatInterface(respond).launch()