Update app.py

app.py

@@ -143,35 +143,51 @@ def filter_models(
     print("Initial df columns:")
     print(df.columns.tolist())
     
-    # Show all models
-    if show_deleted:
-        filtered_df = df
-    else:  # Show only still on the hub models
-        filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True]
+    filtered_df = df.copy()  # Create a copy to avoid modifying the original dataframe
+
+    # Show all models (comment out for debugging)
+    # if not show_deleted:
+    #     filtered_df = filtered_df[filtered_df[AutoEvalColumn.still_on_hub.name] == True]
 
     print(f"After deletion filter: {filtered_df.shape}")
 
+    # Type filter
     type_emoji = [t[0] for t in type_query]
     if 'T' in filtered_df.columns:
-        filtered_df = filtered_df.loc[filtered_df['T'].isin(type_emoji)]
+        filtered_df = filtered_df[filtered_df['T'].isin(type_emoji + ['?'])]  # Include '?' as a valid type
     elif 'Type_Symbol' in filtered_df.columns:
-        filtered_df = filtered_df.loc[filtered_df['Type_Symbol'].isin(type_emoji)]
+        filtered_df = filtered_df[filtered_df['Type_Symbol'].isin(type_emoji + ['?'])]
     else:
         print("Warning: Neither 'T' nor 'Type_Symbol' column found in the dataframe")
-    # filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
-    if 'Unknown' not in precision_query:
-        precision_query.append('Unknown')
-    filtered_df = filtered_df.loc[filtered_df[AutoEvalColumn.precision.name].isin(precision_query)]
-    filtered_df = filtered_df.loc[filtered_df[AutoEvalColumn.add_special_tokens.name].isin(add_special_tokens_query + ["Unknown"])]
-    filtered_df = filtered_df.loc[filtered_df[AutoEvalColumn.num_few_shots.name].isin(num_few_shots_query + ["Unknown"])]
-    print(f"After num_few_shots filter: {filtered_df.shape}")
     
+    print(f"After type filter: {filtered_df.shape}")
+
+    # Precision filter
+    precision_query = precision_query + ['Unknown', '?']
+    filtered_df = filtered_df[filtered_df[AutoEvalColumn.precision.name].isin(precision_query)]
+    print(f"After precision filter: {filtered_df.shape}")
+
+    # Add Special Tokens filter
+    add_special_tokens_query = add_special_tokens_query + ["Unknown"]
+    filtered_df = filtered_df[filtered_df[AutoEvalColumn.add_special_tokens.name].isin(add_special_tokens_query)]
+    print(f"After add_special_tokens filter: {filtered_df.shape}")
+
+    # Num Few Shots filter
+    num_few_shots_query = num_few_shots_query + ["Unknown"]
+    filtered_df = filtered_df[filtered_df[AutoEvalColumn.num_few_shots.name].isin(num_few_shots_query)]
+    print(f"After num_few_shots filter: {filtered_df.shape}")
 
+    # Size filter
     numeric_interval = pd.IntervalIndex(sorted([NUMERIC_INTERVALS[s] for s in size_query]))
-    params_column = pd.to_numeric(df[AutoEvalColumn.params.name], errors="coerce")
-    mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
-    filtered_df = filtered_df.loc[mask]
+    params_column = pd.to_numeric(filtered_df[AutoEvalColumn.params.name], errors="coerce")
+    mask = params_column.apply(lambda x: pd.isna(x) or any(numeric_interval.contains(x)))
+    filtered_df = filtered_df[mask]
     print(f"After size filter: {filtered_df.shape}")
+
+    if filtered_df.empty:
+        print("Warning: Filtered dataframe is empty!")
+        return pd.DataFrame(columns=df.columns)  # Return an empty dataframe with the same columns
+
     print("Filtered dataframe head:")
     print(filtered_df.head())
     print("Column names:")
@@ -180,9 +196,11 @@ def filter_models(
     print(filtered_df.dtypes)
     print("Final filtered dataframe sample:")
     print(filtered_df.head().to_dict('records'))
+
     filtered_df = filtered_df.rename(columns={'Type_Symbol': 'T'})
     print("Filtered DataFrame sample:")
     print(filtered_df.head().to_dict('records'))
+
     filtered_df = filtered_df.astype(str)
     return filtered_df