app.py

import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load Qwen3-0.6B locally with GPU/CPU optimization
model_name = "Qwen/Qwen3-0.6B"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
    device_map="auto" if torch.cuda.is_available() else None
)
model.eval()

def respond(message, history: list[tuple[str, str]], system_message, max_tokens, temperature, top_p):
    # Build chat history
    messages = [{"role": "system", "content": system_message}]
    for user_msg, bot_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if bot_msg:
            messages.append({"role": "assistant", "content": bot_msg})
    messages.append({"role": "user", "content": message})

    # Format messages into a single string for generation
    prompt = ""
    for m in messages:
        prompt += f"{m['role'].capitalize()}: {m['content']}\n"
    prompt += "Assistant:"

    # Tokenize
    input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(model.device)

    # Generate
    output_ids = model.generate(
        input_ids,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True
    )

    # Decode
    output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
    response = output_text[len(prompt):].strip()

    yield response

# Gradio UI
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p"),
    ],
)

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