Handle empty DataFrame case in leaderboard

src/populate.py

@@ -15,17 +15,26 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
     all_data_json = [v.to_dict() for v in raw_data]
 
     df = pd.DataFrame.from_records(all_data_json)
+    
+    # Handle empty DataFrame case
+    if df.empty:
+        # Create empty DataFrame with correct columns
+        df = pd.DataFrame(columns=cols)
+        return df
+    
     # Sort by the first task (EMEA NER) since we don't have an average for NER tasks
     # If no results exist yet, just sort by model name
-    if not df.empty:
-        first_task = list(Tasks)[0]  # emea_ner
-        task_col_name = getattr(AutoEvalColumn, first_task.name).name
-        if task_col_name in df.columns:
-            df = df.sort_values(by=[task_col_name], ascending=False)
-        else:
-            # Fallback to sorting by model name if no task results yet
-            df = df.sort_values(by=[AutoEvalColumn.model.name], ascending=True)
-    df = df[cols].round(decimals=2)
+    first_task = list(Tasks)[0]  # emea_ner
+    task_col_name = getattr(AutoEvalColumn, first_task.name).name
+    if task_col_name in df.columns:
+        df = df.sort_values(by=[task_col_name], ascending=False)
+    else:
+        # Fallback to sorting by model name if no task results yet
+        df = df.sort_values(by=[AutoEvalColumn.model.name], ascending=True)
+    
+    # Only select columns that exist in the DataFrame
+    available_cols = [col for col in cols if col in df.columns]
+    df = df[available_cols].round(decimals=2)
 
     # filter out if any of the benchmarks have not been produced
     df = df[has_no_nan_values(df, benchmark_cols)]