app.py

import threading
import time

import gradio as gr
from huggingface_hub import HfApi
from llama_cpp import Llama

API = HfApi()
LLM = Llama.from_pretrained(
    repo_id="mradermacher/ZEUS-8B-V2-i1-GGUF",
    filename="*Q4_K_M.gguf",
    chat_format="chatml",
    verbose=False
)


def refresh(how_much=43200): # default to 12 hour
  time.sleep(how_much)
  try:
      API.restart_space(repo_id="T145/ZEUS-8B-V2-CHAT")
  except Exception as e:
      print(f"Error while rebooting, trying again... {e}")
      refresh(600) # 10 minutes if any error happens


def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""

    for message in LLM.create_chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        if "choices" not in message:
            continue

        token = message["choices"][0]["delta"]

        if "content" not in token:
            continue

        token = token["content"]

        if token.strip() == "|":
            break

        response += token
        yield response


if __name__ == "__main__":
    demo = gr.ChatInterface(
        fn=respond,
        type="messages",
        additional_inputs=[
            gr.Textbox(value="You are a friendly assistant.", label="System message"),
            gr.Slider(minimum=100, maximum=2048, value=1024, step=2, label="Max new tokens"),
            gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.05, label="Temperature"),
            gr.Slider(
                minimum=0.1,
                maximum=1.0,
                value=0.95,
                step=0.05,
                label="Top-p (nucleus sampling)",
            ),
        ],
    )

    threading.Thread(target=refresh).start()
    demo.launch()