app.py

import gradio as gr
import pandas as pd
import plotly.express as px

def display_table(exam_type):
    if exam_type == "Armenian Exams":
        df = pd.read_csv('unified_exam_results.csv')
        df = df.sort_values(by='Average', ascending=False)
        cols = df.columns.tolist()
        cols.insert(1, cols.pop(cols.index('Average')))
        df = df[cols]
        df.rename(columns={'Armenian language and literature': 'Armenian language\nand literature'}, inplace=True)
        df = df.round(4)
    elif exam_type == "MMLU-Pro-Hy":
        df = pd.read_csv('mmlu_pro_hy_results.csv')
        subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology']
        df['Average'] = df[subject_cols].mean(axis=1)
        df = df.sort_values(by='Average', ascending=False)
        cols = df.columns.tolist()
        cols.remove('Accuracy')
        cols.insert(1, cols.pop(cols.index('Average')))
        cols.append(cols.pop(cols.index('Other')))
        df = df[cols]
        df = df.round(4)
    return df

def create_bar_chart(exam_type, plot_column):
    if exam_type == "Armenian Exams":
        df = pd.read_csv('unified_exam_results.csv')
        df = df.sort_values(by=[plot_column, 'Model'], ascending=[False, True]).reset_index(drop=True)
        x_col = plot_column
        title = f'{plot_column}'
        x_range_max = 20
        def get_label(score):
            if score < 8:
                return "Fail"
            elif 8 <= score <= 18:
                return "Pass"
            else:
                return "Distinction"
        df['Test Result'] = df[plot_column].apply(get_label)
        color_discrete_map = {
            "Fail": "#ff5f56",
            "Pass": "#ffbd2e",
            "Distinction": "#27c93f"
        }
        fig = px.bar(df,
            x=x_col,
            y='Model',
            color=df['Test Result'],
            color_discrete_map=color_discrete_map,
            labels={x_col: 'Score', 'Model': 'Model'},
            title=title,
            orientation='h')
        fig.update_layout(
            xaxis=dict(range=[0, x_range_max]),
            title=dict(text=title, font=dict(size=16)),
            xaxis_title=dict(font=dict(size=12)),
            yaxis_title=dict(font=dict(size=12)),
            yaxis=dict(autorange="reversed"),
            autosize=True
        )
        return fig
    elif exam_type == "MMLU-Pro-Hy":
        df = pd.read_csv('mmlu_pro_hy_results.csv')
        subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology']
        df['Average'] = df[subject_cols].mean(axis=1)
        df = df.sort_values(by=plot_column, ascending=False).reset_index(drop=True)  
        df = df.drop(columns=['Accuracy'])
        x_col = plot_column
        title = f'{plot_column}'
        x_range_max = 1.0
        fig = px.bar(df,
            x=x_col,
            y='Model',
            color=x_col,
            color_continuous_scale='Viridis',
            labels={x_col: 'Accuracy', 'Model': 'Model'},
            title=title,
            orientation='h',
            range_color=[0,1])
        fig.update_layout(
            xaxis=dict(range=[0, x_range_max]),
            title=dict(text=title, font=dict(size=16)),
            xaxis_title=dict(font=dict(size=12)),
            yaxis_title=dict(font=dict(size=12)),
            yaxis=dict(autorange="reversed"),
            autosize=True
        )
        return fig

with gr.Blocks() as app:
    with gr.Tabs():
        with gr.TabItem("Armenian Unified Exams"):
            gr.Markdown("# Armenian Unified Test Exams")
            gr.HTML(f"""
                <div style="font-size: 16px;">
                    This benchmark contains results of various Language Models on Armenian Unified Test Exams for Armenian language and literature, Armenian history and mathematics. The scoring system is a 20-point scale, where 0-8 is a Fail, 8-18 is a Pass, and 18-20 is a Distinction.
                </div>
            """)
            table_output_armenian = gr.DataFrame(value=lambda: display_table("Armenian Exams"))
            plot_column_dropdown = gr.Dropdown(choices=['Average', 'Armenian language and literature', 'Armenian history', 'Mathematics'], value='Average', label='Select Column to Plot')
            plot_output_armenian = gr.Plot(lambda column: create_bar_chart("Armenian Exams", column), inputs=plot_column_dropdown)
        with gr.TabItem("MMLU-Pro-Hy"):
            gr.Markdown("# MMLU-Pro Translated to Armenian (MMLU-Pro-Hy)")
            gr.HTML(f"""
                <div style="font-size: 16px;">
                    This benchmark contains results of various Language Models on the MMLU-Pro benchmark, translated into Armenian. MMLU-Pro is a massive multi-task test in MCQA format. The scores represent accuracy.
                </div>
            """)
            table_output_mmlu = gr.DataFrame(value=lambda: display_table("MMLU-Pro-Hy"))
            subject_cols = ['Average','Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Philosophy', 'Physics', 'Psychology','Other']
            plot_column_dropdown_mmlu = gr.Dropdown(choices=subject_cols, value='Average', label='Select Column to Plot')
            plot_output_mmlu = gr.Plot(lambda column: create_bar_chart("MMLU-Pro-Hy", column), inputs=plot_column_dropdown_mmlu)
app.launch(share=True, debug=True)