app.py

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

# Dictionary to store loaded models and tokenizers
loaded_models = {}

# List of available models (update with your preferred models)
models = ["hanzla/gemma-2b-datascience-instruct-v5", "hanzla/gemma-2b-datascience-instruct-v4.5"]

@spaces.GPU
def load_model(model_name):
    if model_name not in loaded_models:
        print(f"Loading model: {model_name}")
        tokenizer = AutoTokenizer.from_pretrained(model_name)
        model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda" if torch.cuda.is_available() else "cpu")
        loaded_models[model_name] = (model, tokenizer)
    return loaded_models[model_name]

@spaces.GPU
def get_model_response(model_name, message):
    model, tokenizer = load_model(model_name)
    inputs = tokenizer(message, return_tensors="pt").to(model.device)

    outputs = model.generate(**inputs)

    response = tokenizer.decode(outputs[0])
    return response


def chat(message, history, model1, model2):
    response1 = get_model_response(model1, message)
    response2 = get_model_response(model2, message)
    history = history or []
    history.append({"role": "user", "content": message})
    history.append({"role": "assistant", "content": f"{model1}: {response1}\n\n{model2}: {response2}"})
    return history



with gr.Blocks() as demo:
    gr.Markdown("# Hugging Face Model Comparison Chat")

    with gr.Row():
        model1_dropdown = gr.Dropdown(choices=models, label="Model 1", value=models[0])
        model2_dropdown = gr.Dropdown(choices=models, label="Model 2", value=models[1])

    chatbot = gr.Chatbot(label="Chat History", type="messages")
    msg = gr.Textbox(label="Your message")
    clear = gr.Button("Clear")

    with gr.Row():
        upvote = gr.Button("👍 Upvote")
        downvote = gr.Button("👎 Downvote")

    msg.submit(chat, [msg, chatbot, model1_dropdown, model2_dropdown], chatbot)
    clear.click(lambda: [], None, chatbot, queue=False)

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