src/structures/lek_structure.py

import pandas as pd
from pathlib import Path

abs_path = Path(__file__).parent.parent.parent

def replace_models_names(model_name):
    replaces = {'meta-llama': 'meta_llama',
        'epfl-llm':'epfl_llm',
        '01-ai':'01_ai'}
    new_name = model_name.replace('model-', '')
    for k, v in replaces.items():
        if new_name.startswith(k):
            new_name = new_name.replace(k, v)
    new_name = new_name.replace('-','/',1)
    new_name = new_name.replace('_','-',1)
    new_name = f"[{new_name}](https://huggingface.co/{new_name})"
    return new_name

def generate_ORDER_LIST_LEK_and_data_types(json_data):
    ORDER_LIST_LEK = ["model_name", "overall_accuracy"]
    data_types = ["markdown", "number"]

    for key in json_data.keys():
        if key not in ["model_name", "overall_accuracy"]:
            ORDER_LIST_LEK.append(key)
            data_types.append("number")
    ORDER_LIST_LEK[2:] = sorted(ORDER_LIST_LEK[2:])
    return ORDER_LIST_LEK, data_types

def filter_columns_lek(column_choices):
    selected_columns = [col for col in ORDER_LIST_LEK if col in column_choices]
    return LEK_ACCS[selected_columns]

def load_json_data(file_path, ORDER_LIST_LEK):
    LEK_ACCS = pd.read_json(file_path)
    for column in LEK_ACCS.columns:
        if LEK_ACCS[column].apply(type).eq(dict).any():
            LEK_ACCS[column] = LEK_ACCS[column].apply(str)

    LEK_ACCS["model_name"] = LEK_ACCS["model_name"].apply(
        lambda name: replace_models_names(name)
    )

    for column in LEK_ACCS.select_dtypes(include='number').columns:
            LEK_ACCS[column] = LEK_ACCS[column].round(2)
    ordered_columns = [col for col in ORDER_LIST_LEK if col in LEK_ACCS.columns]
    LEK_ACCS = LEK_ACCS[ordered_columns]

    if "Score" in LEK_ACCS.columns:
        LEK_ACCS = LEK_ACCS.sort_values(by="Score", ascending=False)

    return LEK_ACCS


file_path = str(abs_path / "leaderboards/r_lek_report_scores.json")
with open(file_path, 'r', encoding='utf-8') as file:
    sample_data = pd.read_json(file_path).iloc[0].to_dict()  # Load the first row as a dict

ORDER_LIST_LEK, DATA_TYPES_LEK = generate_ORDER_LIST_LEK_and_data_types(sample_data)
LEK_ACCS = load_json_data(file_path, ORDER_LIST_LEK)
# LEK_ACCS = LEK_ACCS.reindex(sorted(LEK_ACCS.columns), axis=1)
COLUMN_HEADERS_LEK = ORDER_LIST_LEK

print(ORDER_LIST_LEK)