Refactor search_leaderboard for improved efficiency and clarity in filtering results

app.py

@@ -44,19 +44,15 @@ retrieval_df = None
 reranking_df = None
 
 def search_leaderboard(df, model_name, columns_to_show, threshold=95):
-    if len(model_name.strip()) == 0:
+    if not model_name.strip():
         return df.loc[:, columns_to_show]
-    
+    search_name = model_name.lower()  # compute once for efficiency
     def calculate_similarity(row):
-        similarity = fuzz.partial_ratio(model_name.lower(), row["Model"].lower())
-        return similarity if similarity >= threshold else 0
-
+        return fuzz.partial_ratio(search_name, row["Model"].lower())
     filtered_df = df.copy()
     filtered_df["similarity"] = filtered_df.apply(calculate_similarity, axis=1)
-    filtered_df = filtered_df[filtered_df["similarity"] > 0].sort_values('similarity', ascending=False)
-    filtered_df = filtered_df.drop('similarity', axis=1)
-    filtered_df = filtered_df.loc[:, columns_to_show]
-
+    filtered_df = filtered_df[filtered_df["similarity"] >= threshold].sort_values('similarity', ascending=False)
+    filtered_df = filtered_df.drop('similarity', axis=1).loc[:, columns_to_show]
     return filtered_df
 
 def retrieval_search_leaderboard(model_name, columns_to_show):
@@ -72,10 +68,12 @@ def main():
     # Prepare retrieval dataframe
     retrieval_df = load_retrieval_results(prepare_for_display=True)
     retrieval_columns_to_show = ["Model", "Web Search Dataset (Overall Score)", "Model Size (in Millions)", "Embedding Dimension", "Max Tokens", "Num Likes"]
+    retrieval_cols = retrieval_df.columns.tolist()  # cache columns
 
     # Prepare reranking dataframe
     reranking_df = load_reranking_results(prepare_for_display=True)
     reranking_columns_to_show = reranking_df.columns.tolist()
+    reranking_cols = reranking_df.columns.tolist()  # cache columns
 
     with gr.Blocks() as demo:
         gr.HTML(HEADER)
@@ -92,7 +90,7 @@ def main():
                             )
                             retrieval_columns_to_show_input = gr.CheckboxGroup(
                                 label="Columns to Show",
-                                choices=retrieval_df.columns.tolist(),
+                                choices=retrieval_cols,  # use cached list
                                 value=retrieval_columns_to_show,
                                 scale=4
                             )
@@ -134,7 +132,7 @@ def main():
                             )
                             reranking_columns_to_show_input = gr.CheckboxGroup(
                                 label="Columns to Show",
-                                choices=reranking_df.columns.tolist(),
+                                choices=reranking_cols,  # use cached list
                                 value=reranking_columns_to_show,
                                 scale=4
                             )
@@ -144,7 +142,7 @@ def main():
                             datatype="markdown",
                             wrap=False,
                             show_fullscreen_button=True,
-                            interactive=False
+                            interactive=False,
                         )
 
                         # Submit the search box and the leaderboard