app.py

import pandas as pd
import gradio as gr
import os
from gradio_rangeslider import RangeSlider

from utils.filter_utils import filter

# MAPS = filter_utils.LANG_MAPPING

# Main Leaderboard containing everything
text_leaderboard = pd.read_csv(os.path.join('src', 'main_df.csv'))
text = "## The range is: {min} to {max}"

# Short leaderboard containing fixed columns
short_leaderboard = text_leaderboard[[
    'model_name', 
    'input_price', 
    'output_price',
    'release_date', 
    'context_size', 
    'average_clemscore',
    'average_latency',
    'parameter_size',
    ]]

## Get Languages
langs = []
for i in range(len(text_leaderboard)):
    lang_splits = text_leaderboard.iloc[i]['languages'].split(',')
    lang_splits = [s.strip() for s in lang_splits]
    langs += lang_splits
langs = list(set(langs))
langs.sort()

## Get input prices
ip_prices = []
op_prices = []
for i in range(len(text_leaderboard)):
    ip_prices.append(text_leaderboard.iloc[i]['input_price'])
    op_prices.append(text_leaderboard.iloc[i]['output_price'])

max_input_price = max(ip_prices)
max_output_price = max(op_prices)


llm_calc_app = gr.Blocks()
with llm_calc_app:

    with gr.Row():
        """
        Main Filters Row
        """
        
    ### Language filter
    with gr.Row():
        lang_dropdown = gr.Dropdown(
            choices=langs,
            value=[],
            multiselect=True,
            label="Select Languages 🕹️"
        )

        clemscore_slider = RangeSlider(
            minimum=0, 
            maximum=100, 
            value=(0, 100), 
            label="Select Clemscore range"
        )

        input_pricing_slider = RangeSlider(
            minimum=0, 
            maximum=max_input_price, 
            value=(0, max_input_price), 
            label="Select Price range /1M input tokens"
        )
       
        output_pricing_slider = RangeSlider(
            minimum=0, 
            maximum=max_output_price, 
            value=(0, max_output_price), 
            label="Select Price range /1M output tokens"
        )
       


    with gr.Row():
        """
        Main Leaderboard Row
        """

        leaderboard_table = gr.Dataframe(
                                value=short_leaderboard,
                                elem_id="text-leaderboard-table",
                                interactive=False,
                                visible=True,
                                height=800
                            )

        
        dummy_leaderboard_table = gr.Dataframe(
                                value=text_leaderboard,
                                elem_id="dummy-leaderboard-table",
                                interactive=False,
                                visible=False
                            )
        
        lang_dropdown.change(
            filter,
            [dummy_leaderboard_table, lang_dropdown, clemscore_slider,
             input_pricing_slider, output_pricing_slider],
            [leaderboard_table],   
            queue=True
        )

        clemscore_slider.change(
            filter,
            [dummy_leaderboard_table, lang_dropdown, clemscore_slider,
             input_pricing_slider, output_pricing_slider],
            [leaderboard_table],
            queue=True
        )

        input_pricing_slider.change(
            filter,
            [dummy_leaderboard_table, lang_dropdown, clemscore_slider,
             input_pricing_slider, output_pricing_slider],
            [leaderboard_table],
            queue=True
        )

        output_pricing_slider.change(
            filter,
            [dummy_leaderboard_table, lang_dropdown, clemscore_slider,
             input_pricing_slider, output_pricing_slider],
            [leaderboard_table],
            queue=True
        )

    llm_calc_app.load()
llm_calc_app.queue()
llm_calc_app.launch()



"""
model_name, input_price, output_price,
multimodality_image,multimodality_multiple_image,multimodality_audio,multimodality_video,
source,licence_name,licence_url,languages,release_date,
parameters_estimated,parameters_actual,

open_weight,context, 

additional_prices_context_caching,
additional_prices_context_storage,
additional_prices_image_input,additional_prices_image_output,additional_prices_video_input,additional_prices_video_output,additional_prices_audio_input,additional_prices_audio_output,clemscore_v1.6.5_multimodal,clemscore_v1.6.5_ascii,clemscore_v1.6,latency_v1.6,latency_v1.6.5_multimodal,latency_v1.6.5_ascii,

average_clemscore,average_latency,parameters

Final list

model_name, input_price, output_price,
multimodality_image,multimodality_multiple_image,multimodality_audio,multimodality_video,
source,licence_name,licence_url,languages,release_date, open_weight,context, average_clemscore,average_latency,parameters


Filter
multimodality_image,multimodality_multiple_image,multimodality_audio,multimodality_video,
licence_name+licence_url, languages, release_date, open_weight

RR
model_name, input_price, output_price,
source, release_date

"""