app.py

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

# Load the model and tokenizer locally
model_name = "mergekit-community/Anti-Qwen2.5-Coder-0.5B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16)

# Ensure the model runs on GPU if available
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Prepare the conversation history for the model
    messages = [f"System: {system_message}"]

    for val in history:
        if val[0]:
            messages.append(f"User: {val[0]}")
        if val[1]:
            messages.append(f"Assistant: {val[1]}")

    messages.append(f"User: {message}")
    context = "\n".join(messages)

    # Tokenize input
    input_ids = tokenizer(context, return_tensors="pt", truncation=True, max_length=2048).input_ids.to(device)

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

    # Decode and yield response
    response = tokenizer.decode(output_ids[0][input_ids.shape[-1]:], skip_special_tokens=True)
    yield response

# Gradio interface setup
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)",
        ),
    ],
)

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