Princeps3007/Jupiter

1import os2import gradio as gr3from huggingface_hub import InferenceClient4 5"""6For more information on `huggingface_hub` Inference API support,7please check the docs:8https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference9"""10 11# Retrieve your token from the environment variable you set in the Space's secrets12token = os.getenv("HF_API_TOKEN")13 14# Instantiate the client using your token15client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=token)16 17def respond(18    message,19    history: list[tuple[str, str]],20    system_message,21    max_tokens,22    temperature,23    top_p,24):25    messages = [{"role": "system", "content": system_message}]26 27    for val in history:28        if val[0]:29            messages.append({"role": "user", "content": val[0]})30        if val[1]:31            messages.append({"role": "assistant", "content": val[1]})32 33    messages.append({"role": "user", "content": message})34 35    response = ""36 37    for msg in client.chat_completion(38        messages,39        max_tokens=max_tokens,40        stream=True,41        temperature=temperature,42        top_p=top_p,43    ):44        token_content = msg.choices[0].delta.content45        response += token_content46        yield response47 48"""49For information on how to customize the ChatInterface,50peruse the gradio docs: https://www.gradio.app/docs/chatinterface51"""52demo = gr.ChatInterface(53    respond,54    additional_inputs=[55        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),56        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),57        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),58        gr.Slider(59            minimum=0.1,60            maximum=1.0,61            value=0.95,62            step=0.05,63            label="Top-p (nucleus sampling)",64        ),65    ],66)67 68if __name__ == "__main__":69    demo.launch()70