app.py

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

# Load your custom model and tokenizer
MODEL_NAME = "Qwen/Qwen2.5-1.5B"  # Replace with your model's Hugging Face repo ID or local path
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)

# Ensure the model is on the CPU
model.to("cpu")

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

    for user_msg, assistant_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if assistant_msg:
            messages.append({"role": "assistant", "content": assistant_msg})

    # Add the latest user message
    messages.append({"role": "user", "content": message})

    # Format the input for the model
    input_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages])

    # Generate a response
    inputs = tokenizer(input_text, return_tensors="pt").to("cpu")  # Move inputs to CPU
    outputs = model.generate(
        inputs.input_ids,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
    )

    # Decode the response
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)

    # Extract only the assistant's response
    # Split the response by "assistant:" and take the last part
    assistant_response = response.split("assistant:")[-1].strip()

    # Remove any repeated history from the response
    # This ensures the response doesn't include the entire conversation
    if "user:" in assistant_response:
        assistant_response = assistant_response.split("user:")[0].strip()

    yield assistant_response


# Create the Gradio interface
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 (nucleus sampling)",
        ),
    ],
)

# Launch the app
if __name__ == "__main__":
    demo.launch()