rank

src/display/utils.py

@@ -52,6 +52,7 @@ auto_eval_column_dict.append(["still_on_hub", ColumnContent, ColumnContent("Avai
 auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)])
 # Dummy column for the search bar (hidden by the custom CSS)
 auto_eval_column_dict.append(["dummy", ColumnContent, ColumnContent("model_name_for_query", "str", False, dummy=True)])
+auto_eval_column_dict.append(["rank", ColumnContent, ColumnContent("rank", "number", True)])
 
 # We use make dataclass to dynamically fill the scores from Tasks
 AutoEvalColumn = make_dataclass("AutoEvalColumn", auto_eval_column_dict, frozen=True)

src/populate.py

@@ -18,6 +18,9 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
     df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
     df = df[cols].round(decimals=2)
 
+    # add column rank to df
+    df["rank"] = df[AutoEvalColumn.average.name].rank(ascending=False, method="min")
+
     # filter out if any of the benchmarks have not been produced
     #df2 = df[has_no_nan_values(df, benchmark_cols)]
     return raw_data, df