app.py

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

# Model configurations
MODELS = {
    "facebook/opt-350m": {
        "name": "Facebook OPT-350M",
        "description": "350M parameter foundation model"
    },
    "EleutherAI/pythia-410m": {
        "name": "EleutherAI Pythia-410M",
        "description": "410M parameter foundation model"
    },
    "openai-community/gpt2-medium": {
        "name": "GPT-2 Medium",
        "description": "355M parameter original GPT-2"
    }
}

# Global storage for models (will be loaded on GPU)
current_model = None
current_tokenizer = None
current_model_name = None

@spaces.GPU
def generate_text(model_name, prompt, max_tokens, temperature, top_p):
    """Generate text using selected model with ZeroGPU"""
    global current_model, current_tokenizer, current_model_name

    try:
        # Load model if not loaded or different model selected
        if current_model is None or current_model_name != model_name:
            current_tokenizer = AutoTokenizer.from_pretrained(model_name)
            current_model = AutoModelForCausalLM.from_pretrained(
                model_name,
                torch_dtype=torch.float16,
                device_map="auto"
            )
            current_model_name = model_name

        inputs = current_tokenizer(prompt, return_tensors="pt").to(current_model.device)

        with torch.no_grad():
            outputs = current_model.generate(
                **inputs,
                max_new_tokens=max_tokens,
                temperature=temperature,
                top_p=top_p,
                do_sample=True,
                pad_token_id=current_tokenizer.eos_token_id
            )

        response = current_tokenizer.decode(outputs[0], skip_special_tokens=True)
        return response

    except Exception as e:
        return f"Error: {str(e)}"

# Gradio interface
with gr.Blocks(title="GPT-2 Class Models Comparison", theme=gr.themes.Soft()) as demo:
    gr.Markdown("# GPT-2 Class Models Comparison (350M-410M)")
    gr.Markdown("Test and compare small foundation models trained on general text")

    with gr.Row():
        model_dropdown = gr.Dropdown(
            choices=list(MODELS.keys()),
            value="facebook/opt-350m",
            label="Select Model",
            info="Choose which 350M-class model to test"
        )

    with gr.Row():
        with gr.Column():
            model_info = gr.Markdown("")

            def update_model_info(model_name):
                info = MODELS[model_name]
                return f"**{info['name']}**\n\n{info['description']}"

            model_dropdown.change(
                fn=update_model_info,
                inputs=[model_dropdown],
                outputs=[model_info]
            )

            # Initialize with default model info
            model_info.value = update_model_info("facebook/opt-350m")

    prompt_box = gr.Textbox(
        label="Prompt",
        placeholder="Type your prompt here...",
        lines=5
    )

    with gr.Row():
        max_tokens = gr.Slider(
            minimum=50,
            maximum=500,
            value=200,
            step=10,
            label="Max New Tokens"
        )
        temperature = gr.Slider(
            minimum=0.1,
            maximum=2.0,
            value=0.7,
            step=0.1,
            label="Temperature"
        )
        top_p = gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.9,
            step=0.05,
            label="Top P"
        )

    generate_btn = gr.Button("Generate", variant="primary", size="lg")

    output_box = gr.Textbox(
        label="Generated Text",
        lines=10,
        interactive=False
    )

    generate_btn.click(
        fn=generate_text,
        inputs=[model_dropdown, prompt_box, max_tokens, temperature, top_p],
        outputs=[output_box]
    )

    gr.Markdown("---")
    gr.Markdown("### Example Prompts:")
    gr.Markdown("""
    - General: "Once upon a time in a distant galaxy"
    - Technical: "The main advantages of Python programming are"
    - Medical: "Question: What are the indications for aspirin in cardiac patients?\nAnswer:"
    """)

demo.launch()