f

app.py

@@ -79,8 +79,8 @@ def style_df(df: pd.DataFrame) -> Styler:
     leaderboard_df_styled = df.style.background_gradient(cmap="RdYlGn")
 
     inverted_colors_columns = ['rank']
-    if "poleval2018_task3_test_10k" in df.columns:
-        inverted_colors_columns.append("poleval2018_task3_test_10k")
+    # if "poleval2018_task3_test_10k" in df.columns:
+    #     inverted_colors_columns.append("poleval2018_task3_test_10k")
     if '#Params (B)' in df.columns:
         inverted_colors_columns.append("#Params (B)")
 
@@ -88,7 +88,7 @@ def style_df(df: pd.DataFrame) -> Styler:
     rounding = {'#Params (B)': "{:.1f}", 'rank': "{:.0f}"}
     for task in Tasks:
         rounding[task.value.col_name] = "{:.2f}"
-    for column_name in ["Average ⬆️", "Avg g", "Avg mc", "Average old", "Avg RAG"]:
+    for column_name in ["Average ⬆️"]:
         rounding[column_name] = "{:.2f}"
     leaderboard_df_styled = leaderboard_df_styled.format(rounding)
     return leaderboard_df_styled

src/leaderboard/read_evals.py

@@ -387,8 +387,8 @@ def get_raw_eval_results(results_path: str, requests_path: str, metadata) -> lis
     model_result_filepaths = []
 
     for root, _, files in os.walk(results_path):
-        print(root)
-        # if '_polish_pes_' not in root: continue
+        if '_polish_pes_' not in root: continue
+
         # We should only have json files in model results
         if len(files) == 0 or any([not f.endswith(".json") for f in files]):
             continue