lowercase models

app.py

@@ -225,6 +225,11 @@ def init_leaderboard(dataframe, visible_columns=None):
         dataframe = pd.DataFrame(columns=columns)
         logger.warning("Initializing empty leaderboard")
 
+    # Lowercase model_name for display
+    if "model_name" in dataframe.columns:
+        dataframe = dataframe.copy()
+        dataframe["model_name"] = dataframe["model_name"].str.lower()
+
     # print("\n\n", "dataframe", dataframe, "--------------------------------\n\n")
 
     # Determine which columns to display
@@ -588,36 +593,21 @@ def create_performance_plot(
     if df.empty:
         return go.Figure()
 
-    # Filter for selected models
+    # Lowercase model_name in df and selected_models
+    df = df.copy()
+    df["model_name"] = df["model_name"].str.lower()
+    selected_models = [m.lower() for m in selected_models]
     df = df[df["model_name"].isin(selected_models)]
-
-    # Get the relevant metric columns
     metric_cols = [col for col in df.columns if metric in col]
-
-    # Create figure
     fig = go.Figure()
-
-    # Custom colors for different models
-    colors = [
-        "#8FCCCC",
-        "#C2A4B6",
-        "#98B4A6",
-        "#B68F7C",
-    ]  # Pale Cyan, Pale Pink, Pale Green, Pale Orange
-
-    # Add traces for each model
+    colors = ["#8FCCCC", "#C2A4B6", "#98B4A6", "#B68F7C"]
     for idx, model in enumerate(selected_models):
         model_data = df[df["model_name"] == model]
         if not model_data.empty:
             values = model_data[metric_cols].values[0].tolist()
-            # Add the first value again at the end to complete the polygon
             values = values + [values[0]]
-
-            # Clean up test type names
             categories = [col.replace(f"_{metric}", "") for col in metric_cols]
-            # Add the first category again at the end to complete the polygon
             categories = categories + [categories[0]]
-
             fig.add_trace(
                 go.Scatterpolar(
                     r=values,
@@ -627,8 +617,6 @@ def create_performance_plot(
                     fill="toself",
                 )
             )
-
-    # Update layout with all settings at once
     fig.update_layout(
         paper_bgcolor="#000000",
         plot_bgcolor="#000000",
@@ -663,7 +651,6 @@ def create_performance_plot(
             font={"color": "#ffffff"},
         ),
     )
-
     return fig
 
 
@@ -674,7 +661,7 @@ def update_model_choices(version):
     df = get_leaderboard_df(version=version)
     if df.empty:
         return []
-    return sorted(df["model_name"].unique().tolist())
+    return sorted(df["model_name"].str.lower().unique().tolist())
 
 
 def update_visualization(selected_models, selected_category, selected_metric, version):
@@ -744,7 +731,7 @@ with demo:
                     )
                     model_type_filter = gr.Dropdown(
                         choices=[
-                            t.to_str(" : ") for t in ModelType if t != ModelType.Unknown
+                            t.to_str("-") for t in ModelType if t != ModelType.Unknown
                         ],
                         label="Access Type",
                         multiselect=True,
@@ -981,15 +968,10 @@ with demo:
                             df = get_leaderboard_df(version=version)
                             if df.empty:
                                 return []
-                            # Return list of tuples (model_name, mode)
-                            return sorted(
-                                [
-                                    f"{row['model_name']} [{row['mode']}]"
-                                    for _, row in df.drop_duplicates(
-                                        subset=["model_name", "mode"]
-                                    ).iterrows()
-                                ]
-                            )
+                            return sorted([
+                                f"{str(row['model_name']).lower()} [{row['mode']}]"
+                                for _, row in df.drop_duplicates(subset=["model_name", "mode"]).iterrows()
+                            ])
 
                         model_mode_selector = gr.Dropdown(
                             choices=get_model_mode_choices(CURRENT_VERSION),
@@ -1032,27 +1014,23 @@ with demo:
                     df = (
                         get_leaderboard_df(version=version)
                         if selected_category == "All Results"
-                        else get_category_leaderboard_df(
-                            selected_category, version=version
-                        )
+                        else get_category_leaderboard_df(selected_category, version=version)
                     )
                     if df.empty:
                         return go.Figure()
-                    # Parse selected_model_modes into model_name and mode
+                    df = df.copy()
+                    df["model_name"] = df["model_name"].str.lower()
                     selected_pairs = [s.rsplit(" [", 1) for s in selected_model_modes]
                     selected_pairs = [
-                        (name.strip(), mode.strip("] "))
+                        (name.strip().lower(), mode.strip("] "))
                         for name, mode in selected_pairs
                     ]
                     mask = df.apply(
-                        lambda row: (row["model_name"], str(row["mode"]))
-                        in selected_pairs,
+                        lambda row: (row["model_name"], str(row["mode"])) in selected_pairs,
                         axis=1,
                     )
                     filtered_df = df[mask]
-                    metric_cols = [
-                        col for col in filtered_df.columns if selected_metric in col
-                    ]
+                    metric_cols = [col for col in filtered_df.columns if selected_metric in col]
                     fig = go.Figure()
                     colors = ["#8FCCCC", "#C2A4B6", "#98B4A6", "#B68F7C"]
                     for idx, (model_name, mode) in enumerate(selected_pairs):
@@ -1063,10 +1041,7 @@ with demo:
                         if not model_data.empty:
                             values = model_data[metric_cols].values[0].tolist()
                             values = values + [values[0]]
-                            categories = [
-                                col.replace(f"_{selected_metric}", "")
-                                for col in metric_cols
-                            ]
+                            categories = [col.replace(f"_{selected_metric}", "") for col in metric_cols]
                             categories = categories + [categories[0]]
                             fig.add_trace(
                                 go.Scatterpolar(
@@ -1175,7 +1150,7 @@ with demo:
                         )
                         model_type = gr.Dropdown(
                             choices=[
-                                t.to_str(" : ")
+                                t.to_str("-")
                                 for t in ModelType
                                 if t != ModelType.Unknown
                             ],

src/display/utils.py

@@ -24,7 +24,7 @@ class ModelType(Enum):
     ClosedSource = auto()
     API = auto()
 
-    def to_str(self, separator: str = " ") -> str:
+    def to_str(self, separator: str = "-") -> str:
         """Convert enum to string with separator."""
         if self == ModelType.Unknown:
             return "Unknown"
@@ -44,7 +44,7 @@ class GuardModelType(str, Enum):
     OPENAI_MODERATION = "openai_moderation"
     LLM_REGEXP = "llm_regexp"
     LLM_SO = "llm_so"
-    WC_GUARD = "wc_guard"
+    WHITECIRCLE_GUARD = "whitecircle_guard"
 
     def __str__(self):
         """String representation of the guard model type."""