app.py

import os
os.system('pip install minijinja')
import gradio as gr
from huggingface_hub import InferenceClient
import torch
import spaces
# Initialize the client with your model
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")  # Replace with your model's name or endpoint

default_system = 'You are a helpful assistant'

@spaces.GPU
def respond(message, history, system_message, max_tokens, temperature, top_p):
    messages = [{"role": "system", "content": system_message}]
    for user, assistant in history:
        messages.append({"role": "user", "content": user})
        messages.append({"role": "assistant", "content": assistant})
    messages.append({"role": "user", "content": message})

    response = ""
    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

def clear_session():
    return "", []

def modify_system_session(system):
    if not system:
        system = default_system
    return system, system, []

with gr.Blocks() as demo:
    gr.Markdown("<h1 style='text-align: center;'>LLM.C 1.5B Chat Demo 🤖</h1>")
    
    with gr.Row():
        with gr.Column(scale=3):
            system_input = gr.Textbox(value=default_system, lines=1, label='System Prompt')
        with gr.Column(scale=3):
            modify_system = gr.Button("🛠️ Set system prompt and clear history")
    
    system_state = gr.Textbox(value=default_system, visible=False)
    chatbot = gr.Chatbot(label='LLM.C Chat')
    message = gr.Textbox(lines=1, label='Your message')
    
    with gr.Row():
        clear_history = gr.Button("🧹 Clear history")
        submit = gr.Button("🚀 Send")
    
    with gr.Accordion("Advanced Settings", open=False):
        max_tokens = gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max New Tokens")
        temperature = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
        top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (Nucleus Sampling)")

    # Set up event handlers
    message.submit(respond, inputs=[message, chatbot, system_state, max_tokens, temperature, top_p], outputs=[chatbot])
    submit.click(respond, inputs=[message, chatbot, system_state, max_tokens, temperature, top_p], outputs=[chatbot])
    clear_history.click(fn=clear_session, inputs=[], outputs=[message, chatbot])
    modify_system.click(fn=modify_system_session, inputs=[system_input], outputs=[system_state, system_input, chatbot])

    gr.Markdown(
        """
        ## About LLM.C
        Some stuff about llm.c + link to the github repo
        """
    )

if __name__ == "__main__":
    demo.launch()