src/populate.py

import json
from pathlib import Path

import pandas as pd

from src.display.formatting import has_no_nan_values, make_clickable_model
from src.display.utils import AutoEvalColumn, EvalQueueColumn
from src.leaderboard.read_evals import get_raw_eval_results


def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
    """Creates a dataframe from all the individual experiment results"""
    raw_data = get_raw_eval_results(results_path)
    all_data_json = [v.to_dict() for v in raw_data]

    df = pd.DataFrame.from_records(all_data_json)
    df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
    df = df[cols].round(decimals=2)

    return df


def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
    """Creates the different dataframes for the evaluation queues requestes"""
    all_evals = []

    for file_path in Path(save_path).rglob("requests_*.json"):
        with open(file_path) as fp:
            data = json.load(fp)["leaderboard"]

        data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
        data[EvalQueueColumn.revision.name] = data.get("revision", "main")

        all_evals.append(data)

    pending_list = [e for e in all_evals if e["status"] in ["PENDING", "RERUN"]]
    running_list = [e for e in all_evals if e["status"] == "RUNNING"]
    finished_list = [e for e in all_evals if e["status"].startswith("FINISHED") or e["status"] == "PENDING_NEW_EVAL"]
    df_pending = pd.DataFrame.from_records(pending_list, columns=cols)
    df_running = pd.DataFrame.from_records(running_list, columns=cols)
    df_finished = pd.DataFrame.from_records(finished_list, columns=cols)
    return df_finished[cols], df_running[cols], df_pending[cols]