Update app.py

app.py

@@ -129,12 +129,17 @@ def filter_queries(query: str, filtered_df: pd.DataFrame):
 def filter_models(
     df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, add_special_tokens_query: list, num_few_shots_query: list, show_deleted: bool, show_merges: bool, show_flagged: bool
 ) -> pd.DataFrame:
+    print(f"filter_models called with: type_query={type_query}, size_query={size_query}, precision_query={precision_query}")
+    print(f"Initial df shape: {df.shape}")
+    
     # 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]
 
+    print(f"After deletion filter: {filtered_df.shape}")
+
     #if not show_merges:
     #    filtered_df = filtered_df[filtered_df[AutoEvalColumn.merged.name] == False]
 
@@ -143,15 +148,30 @@ def filter_models(
 
     type_emoji = [t[0] for t in type_query]
     filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
-    filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])]
-    filtered_df = filtered_df.loc[df[AutoEvalColumn.add_special_tokens.name].isin(add_special_tokens_query)]
-    filtered_df = filtered_df.loc[df[AutoEvalColumn.num_few_shots.name].isin(num_few_shots_query)]
+    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}")
     
 
     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]
+    print(f"After size filter: {filtered_df.shape}")
+    print("Filtered dataframe head:")
+    print(filtered_df.head())
+    print("Column names:")
+    print(filtered_df.columns.tolist())
+    print("Column data types:")
+    print(filtered_df.dtypes)
+    filtered_df = filtered_df.rename(columns={'T': 'Type_Symbol'})
+    print("Final filtered dataframe columns:")
+    print(filtered_df.columns.tolist())
+    print("Final filtered dataframe sample:")
+    print(filtered_df.head().to_dict('records'))
     return filtered_df
 
 leaderboard_df = filter_models(leaderboard_df, [t.to_str(" : ") for t in ModelType], list(NUMERIC_INTERVALS.keys()), [i.value.name for i in Precision], [i.value.name for i in AddSpecialTokens], [i.value.name for i in NumFewShots], False, False, False)