app.py

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

# Load DialoGPT model and tokenizer
model_name = "microsoft/DialoGPT-large"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Respond function for Gradio interface
def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Format the chat history for the DialoGPT model using the 'messages' format
    conversation = [{"role": "system", "content": system_message}]
    for user_msg, bot_msg in history:
        if user_msg:
            conversation.append({"role": "user", "content": user_msg})
        if bot_msg:
            conversation.append({"role": "assistant", "content": bot_msg})
    conversation.append({"role": "user", "content": message})

    # Tokenize input and generate response
    inputs = tokenizer.encode(" ".join([msg["content"] for msg in conversation]), return_tensors="pt")
    outputs = model.generate(
        inputs,
        max_length=max_tokens,
        temperature=temperature,
        top_p=top_p,
        pad_token_id=tokenizer.eos_token_id,
    )
    response = tokenizer.decode(outputs[:, inputs.shape[-1] :][0], skip_special_tokens=True)
    return response

# Gradio Chat Interface
demo = gr.Interface(
    fn=respond,
    inputs=[
        gr.Textbox(label="Message"),
        gr.State(),  # For history
        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)",
        ),
    ],
    outputs="text",
)

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