src/populate.py

import pandas as pd
import os

from src.display.utils import AutoEvalColumn, ModelType
from src.about import Tasks

def get_leaderboard_df(cols: list, benchmark_cols: list) -> pd.DataFrame:
    """Creates a dataframe from the static results.csv"""
    csv_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), "results.csv")
    df = pd.read_csv(csv_path)

    # Add model_type_symbol based on model_type
    def get_type_with_symbol(mtype):
        return ModelType.from_str(str(mtype)).to_str()
        
    df["model_type"] = df["model_type"].apply(get_type_with_symbol)
    
    # Sort by pass_rate
    if "pass_rate" in df.columns:
        df = df.sort_values(by=["pass_rate"], ascending=False)
        
    # Rename columns to match the expected names in AutoEvalColumn
    rename_map = {
        "model": AutoEvalColumn.model.name,
        "model_type": AutoEvalColumn.model_type.name,
        "params": AutoEvalColumn.params.name,
        "license": AutoEvalColumn.license.name,
    }
    
    # Add tasks to rename map
    for task in Tasks:
        # task.value.benchmark is the column name in csv (e.g., "pass_rate")
        # task.value.col_name is the display name (e.g., "Pass Rate (%)")
        rename_map[task.value.benchmark] = task.value.col_name
        
    df = df.rename(columns=rename_map)
        
    # Ensure all required columns exist, fill missing with NaN
    for col in cols:
        if col not in df.columns:
            df[col] = None

    df = df[cols].round(decimals=2)
    return df