debug

app.py

@@ -46,7 +46,6 @@ from src.submission.submit import add_new_eval
 
 
 
-
 from gradio_leaderboard import SelectColumns, Leaderboard
 import pandas as pd
 from typing import List, Dict, Union, Optional, Any
@@ -54,43 +53,31 @@ from dataclasses import fields
 
 class SmartSelectColumns(SelectColumns):
     """
-    Enhanced SelectColumns component for gradio_leaderboard with dynamic column filtering.
+    Enhanced SelectColumns component with basic filtering functionality.
     """
     def __init__(
         self,
         benchmark_keywords: Optional[List[str]] = None,
         model_keywords: Optional[List[str]] = None,
-        column_mapping: Optional[Dict[str, str]] = None,
         initial_selected: Optional[List[str]] = None,
         **kwargs
     ):
         """
-        Initialize SmartSelectColumns with dynamic filtering.
+        Initialize SmartSelectColumns with minimal configuration.
         
         Args:
-            benchmark_keywords: List of benchmark names to filter by (e.g., ["ioi", "mcqa"])
-            model_keywords: List of model names to filter by (e.g., ["llama3", "qwen2_5"])
-            column_mapping: Dict mapping actual column names to display names
+            benchmark_keywords: List of benchmark names to filter by
+            model_keywords: List of model names to filter by
             initial_selected: List of columns to show initially
         """
         super().__init__(**kwargs)
         self.benchmark_keywords = benchmark_keywords or []
         self.model_keywords = model_keywords or []
-        self.column_mapping = column_mapping or {}
-        self.reverse_mapping = {v: k for k, v in self.column_mapping.items()} if column_mapping else {}
         self.initial_selected = initial_selected or []
-        
-    def preprocess_value(self, x: List[str]) -> List[str]:
-        """Transform selected display names back to actual column names."""
-        return [self.reverse_mapping.get(col, col) for col in x]
-    
-    def postprocess_value(self, y: List[str]) -> List[str]:
-        """Transform actual column names to display names."""
-        return [self.column_mapping.get(col, col) for col in y]
 
     def get_filtered_groups(self, df: pd.DataFrame) -> Dict[str, List[str]]:
         """
-        Dynamically create column groups based on keywords.
+        Create column groups based on simple substring matching.
         """
         filtered_groups = {}
         
@@ -102,10 +89,7 @@ class SmartSelectColumns(SelectColumns):
             ]
             if matching_cols:
                 group_name = f"Benchmark group for {benchmark}"
-                filtered_groups[group_name] = [
-                    self.column_mapping.get(col, col) 
-                    for col in matching_cols
-                ]
+                filtered_groups[group_name] = matching_cols
         
         # Create model groups
         for model in self.model_keywords:
@@ -115,10 +99,7 @@ class SmartSelectColumns(SelectColumns):
             ]
             if matching_cols:
                 group_name = f"Model group for {model}"
-                filtered_groups[group_name] = [
-                    self.column_mapping.get(col, col) 
-                    for col in matching_cols
-                ]
+                filtered_groups[group_name] = matching_cols
         
         return filtered_groups
 
@@ -128,13 +109,8 @@ class SmartSelectColumns(SelectColumns):
     ) -> Dict:
         """Update component with new values."""
         if isinstance(value, pd.DataFrame):
-            # Get all column names and convert to display names
-            choices = [self.column_mapping.get(col, col) for col in value.columns]
-            
-            # Use initial selection or default columns
+            choices = list(value.columns)
             selected = self.initial_selected if self.initial_selected else choices
-            
-            # Get dynamically filtered groups
             filtered_cols = self.get_filtered_groups(value)
             
             return {
@@ -143,13 +119,11 @@ class SmartSelectColumns(SelectColumns):
                 "filtered_cols": filtered_cols
             }
         
-        # Handle fields object
         if hasattr(value, '__dataclass_fields__'):
             field_names = [field.name for field in fields(value)]
-            choices = [self.column_mapping.get(name, name) for name in field_names]
             return {
-                "choices": choices,
-                "value": self.initial_selected if self.initial_selected else choices
+                "choices": field_names,
+                "value": self.initial_selected if self.initial_selected else field_names
             }
             
         return super().update(value)
@@ -366,6 +340,7 @@ def init_leaderboard_mib_subgraph(dataframe, track):
     # Important: We need to rename our DataFrame columns to match display names
     renamed_df = dataframe.rename(columns=display_mapping)
 
+    # Original code 
     # return Leaderboard(
     #     value=renamed_df,  # Use DataFrame with display names
     #     datatype=[c.type for c in fields(AutoEvalColumn_mib_subgraph)],
@@ -378,51 +353,80 @@ def init_leaderboard_mib_subgraph(dataframe, track):
     #     interactive=False,
     # )
     
-    # Complete column groups for both benchmarks and models
-    # Define keywords for filtering
-    benchmark_keywords = ["ioi", "mcqa", "arithmetic_addition", "arithmetic_subtraction", "arc_easy", "arc_challenge"]
+
+
+
+    # # Complete column groups for both benchmarks and models
+    # # Define keywords for filtering
+    # benchmark_keywords = ["ioi", "mcqa", "arithmetic_addition", "arithmetic_subtraction", "arc_easy", "arc_challenge"]
+    # model_keywords = ["qwen2_5", "gpt2", "gemma2", "llama3"]
+
+    # # Optional: Define display names
+    # mappings = {
+    #     "ioi_llama3": "IOI (LLaMA-3)",
+    #     "ioi_qwen2_5": "IOI (Qwen-2.5)",
+    #     "ioi_gpt2": "IOI (GPT-2)",
+    #     "ioi_gemma2": "IOI (Gemma-2)",
+    #     "mcqa_llama3": "MCQA (LLaMA-3)",
+    #     "mcqa_qwen2_5": "MCQA (Qwen-2.5)",
+    #     "mcqa_gemma2": "MCQA (Gemma-2)",
+    #     "arithmetic_addition_llama3": "Arithmetic Addition (LLaMA-3)",
+    #     "arithmetic_subtraction_llama3": "Arithmetic Subtraction (LLaMA-3)",
+    #     "arc_easy_llama3": "ARC Easy (LLaMA-3)",
+    #     "arc_easy_gemma2": "ARC Easy (Gemma-2)",
+    #     "arc_challenge_llama3": "ARC Challenge (LLaMA-3)",
+    #     "eval_name": "Evaluation Name",
+    #     "Method": "Method",
+    #     "Average": "Average Score"
+    # }
+    # # mappings = {}
+
+    # # Create SmartSelectColumns instance
+    # smart_columns = SmartSelectColumns(
+    #     benchmark_keywords=benchmark_keywords,
+    #     model_keywords=model_keywords,
+    #     column_mapping=mappings,
+    #     initial_selected=["Method", "Average"]
+    # )
+
+    # print("\nDebugging DataFrame columns:", renamed_df.columns.tolist())
+
+    # # Create Leaderboard
+    # leaderboard = Leaderboard(
+    #     value=renamed_df,
+    #     datatype=[c.type for c in fields(AutoEvalColumn_mib_subgraph)],
+    #     select_columns=smart_columns,
+    #     search_columns=["Method"],
+    #     hide_columns=[],
+    #     interactive=False
+    # )
+    # print(f"Successfully created leaderboard.")
+    # return leaderboard
+
+    print("\nDebugging DataFrame columns:", dataframe.columns.tolist())
+
+    # Define simple keywords for filtering
+    benchmark_keywords = ["ioi", "mcqa", "arithmetic", "arc"]
     model_keywords = ["qwen2_5", "gpt2", "gemma2", "llama3"]
 
-    # Optional: Define display names
-    mappings = {
-        "ioi_llama3": "IOI (LLaMA-3)",
-        "ioi_qwen2_5": "IOI (Qwen-2.5)",
-        "ioi_gpt2": "IOI (GPT-2)",
-        "ioi_gemma2": "IOI (Gemma-2)",
-        "mcqa_llama3": "MCQA (LLaMA-3)",
-        "mcqa_qwen2_5": "MCQA (Qwen-2.5)",
-        "mcqa_gemma2": "MCQA (Gemma-2)",
-        "arithmetic_addition_llama3": "Arithmetic Addition (LLaMA-3)",
-        "arithmetic_subtraction_llama3": "Arithmetic Subtraction (LLaMA-3)",
-        "arc_easy_llama3": "ARC Easy (LLaMA-3)",
-        "arc_easy_gemma2": "ARC Easy (Gemma-2)",
-        "arc_challenge_llama3": "ARC Challenge (LLaMA-3)",
-        "eval_name": "Evaluation Name",
-        "Method": "Method",
-        "Average": "Average Score"
-    }
-    # mappings = {}
-
-    # Create SmartSelectColumns instance
+    # Create SmartSelectColumns with minimal configuration
     smart_columns = SmartSelectColumns(
         benchmark_keywords=benchmark_keywords,
         model_keywords=model_keywords,
-        column_mapping=mappings,
         initial_selected=["Method", "Average"]
     )
 
-    print("\nDebugging DataFrame columns:", renamed_df.columns.tolist())
-
-    # Create Leaderboard
+    # Create and return the leaderboard
+    print("\nCreating leaderboard...")
     leaderboard = Leaderboard(
-        value=renamed_df,
+        value=dataframe,
         datatype=[c.type for c in fields(AutoEvalColumn_mib_subgraph)],
         select_columns=smart_columns,
         search_columns=["Method"],
         hide_columns=[],
         interactive=False
     )
-    print(f"Successfully created leaderboard.")
+    print("Leaderboard created successfully")
     return leaderboard
 
 
@@ -430,9 +434,6 @@ def init_leaderboard_mib_subgraph(dataframe, track):
 
 
 
-
-
-
 # def init_leaderboard_mib_subgraph(dataframe, track):
 #     """Initialize the subgraph leaderboard with group-based column selection."""
 #     if dataframe is None or dataframe.empty: