debug

src/populate.py

@@ -109,13 +109,43 @@ def aggregate_methods(df: pd.DataFrame) -> pd.DataFrame:
     
     return aggregated_df
 
+# def create_intervention_averaged_df(df: pd.DataFrame) -> pd.DataFrame:
+#     """Creates a DataFrame where columns are model_task and cells are averaged over interventions"""
+#     df_copy = df.copy()
+    
+#     # Remove the Method column and eval_name if present
+#     columns_to_drop = ['Method', 'eval_name']
+#     df_copy = df_copy.drop(columns=[col for col in columns_to_drop if col in df_copy.columns])
+    
+#     # Group columns by model_task
+#     model_task_groups = {}
+#     for col in df_copy.columns:
+#         model_task = '_'.join(col.split('_')[:2])  # Get model_task part
+#         if model_task not in model_task_groups:
+#             model_task_groups[model_task] = []
+#         model_task_groups[model_task].append(col)
+    
+#     # Create new DataFrame with averaged intervention scores
+#     averaged_df = pd.DataFrame({
+#         model_task: df_copy[cols].mean(axis=1).round(3)
+#         for model_task, cols in model_task_groups.items()
+#     })
+    
+#     return averaged_df
+
 def create_intervention_averaged_df(df: pd.DataFrame) -> pd.DataFrame:
     """Creates a DataFrame where columns are model_task and cells are averaged over interventions"""
     df_copy = df.copy()
     
-    # Remove the Method column and eval_name if present
-    columns_to_drop = ['Method', 'eval_name']
-    df_copy = df_copy.drop(columns=[col for col in columns_to_drop if col in df_copy.columns])
+    # Store Method column if it exists
+    method_col = None
+    if 'Method' in df_copy.columns:
+        method_col = df_copy['Method']
+        df_copy = df_copy.drop('Method', axis=1)
+    
+    # Remove eval_name if present
+    if 'eval_name' in df_copy.columns:
+        df_copy = df_copy.drop('eval_name', axis=1)
     
     # Group columns by model_task
     model_task_groups = {}
@@ -131,6 +161,10 @@ def create_intervention_averaged_df(df: pd.DataFrame) -> pd.DataFrame:
         for model_task, cols in model_task_groups.items()
     })
     
+    # Add Method column back
+    if method_col is not None:
+        averaged_df.insert(0, 'Method', method_col)
+    
     return averaged_df
 
 # def get_leaderboard_df_mib_causalgraph(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame: