import gradio as gr
import pandas as pd

# Define benchmark data
benchmark_data = {
    'Model': [
        'IlyaGusev/saiga_llama3_8b',              # LLaMA3
        'Vikhrmodels/Vikhr-Nemo-12B',       # Vikhr
        'TinyLLaMA/TinyLlama-1.1B',         # TinyLLaMA
        'mistralai/Mistral-Nemo-Instruct-2407',     # Mistral
        'Vikhrmodels/Vikhr-Qwen-2.5-0.5b-Instruct'                       # Qwen
    ],
    'Creativity Score': [
        37.75,  # LLaMA3
        46.00,  # Vikhr
        6.50,   # TinyLLaMA
        23.75,  # Mistral
        8.25    # Qwen
    ],
    'Diversity Score': [
        49.50,  # LLaMA3
        52.00,  # Vikhr
        14.50,  # TinyLLaMA
        38.50,  # Mistral
        15.55   # Qwen
    ],
    'Relevance Score': [
        79.25,  # LLaMA3
        87.50,  # Vikhr
        18.50,  # TinyLLaMA
        76.75,  # Mistral
        34.25   # Qwen
    ],
    'Average Score': [
        55.50,  # LLaMA3
        61.83,  # Vikhr
        13.17,  # TinyLLaMA
        46.33,  # Mistral
        19.35   # Qwen
    ]
}

def display_results():
    df = pd.DataFrame(benchmark_data)
    return df

# Create Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Russian Language Model Benchmark Results")
    
    # Add dataframe output
    output = gr.DataFrame(
        headers=list(benchmark_data.keys()),
        interactive=False
    )
    
    refresh_btn = gr.Button("Show Results")
    refresh_btn.click(fn=display_results, outputs=output)

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