Update populate to add rank and model information (#2)

- Update populate to add rank and model information (5f986f5a6e02e7902fbdb271bd9958ee9296a14d)


Co-authored-by: Jack Cole <jcole1@users.noreply.huggingface.co>

src/populate.py

@@ -1,8 +1,9 @@
 import json
 import os
-
+import numpy as np
 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
@@ -11,15 +12,21 @@ 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, requests_path)
-    all_data_json = [v.to_dict() for v in raw_data]
+    for result in raw_data:
+        result.average = np.mean(list(result.results.values()))
+    sorted_results = sorted(raw_data, key=lambda r: r.average, reverse=True)
+    # ranks = [rank+1 for rank, value in enumerate(sorted_results)]
+    # rank = [rank+1 for rank, value in enumerate(average)]
+    all_data_json = [v.to_dict(i+1) for i, v in enumerate(raw_data)]
 
     df = pd.DataFrame.from_records(all_data_json)
-    df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
+    # df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
     df = df[cols].round(decimals=2)
 
     # filter out if any of the benchmarks have not been produced
     df = df[has_no_nan_values(df, benchmark_cols)]
-    return raw_data, df
+    print(df)
+    return df
 
 
 def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
@@ -55,4 +62,4 @@ def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
     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]
+    return df_finished[cols], df_running[cols], df_pending[cols]