Update src/streamlit_app.py

src/streamlit_app.py

@@ -172,20 +172,19 @@ def create_bar_chart(df, view_type):
             color_continuous_scale=px.colors.diverging.Fall,
             orientation="v",
         )
-        
         fig.update_layout(
             xaxis_title_text="Model",
             yaxis_title_text="Score (Lower is better)",
             title_text="",
             font=dict(size=15),
             xaxis_tickangle=-45,
-            bargap=0.2, # Increase gap for slimmer bars
-            height=600, # Set the height of the plot
+            bargap=0.2,
+            height=600,
             showlegend=False,
             margin=dict(
                 l=80,   # Left
                 r=0,    # Right
-                b=80,  # Bottom
+                b=80,   # Bottom
                 t=70,   # Top
                 pad=0   # Padding
             ),
@@ -199,48 +198,87 @@ def create_bar_chart(df, view_type):
             texttemplate="%{y:.2f}",
             textposition="outside"
         )
-
-
-        # TODO
-        # fig = go.Figure(data=[
-        #     go.Bar(
-        #         x=df['Model'],
-        #         y=df['Total Score'],
-        #         orientation='v',
-        #         marker_color=px.colors.sequential.Blues,
-        #         text=df['Total Score'].round(1),
-        #         textposition='outside',
-        #     )
-        # ])
-        # fig.update_layout(
-        #     title="Model Performance - Total Score",
-        #     xaxis_title="Model",
-        #     yaxis_title="Score",
-        #     yaxis_range=[0, 100],
-        #     height=500,
-        # )
     
     elif view_type == "Per Embodiment":
-        # embodiment_cols = [col for col in df.columns if col.startswith('Embodiment-')]
-        fig = go.Figure()
-        # for col in embodiment_cols:
-        #     fig.add_trace(go.Bar(
-        #         name=col.replace('Embodiment-', ''),
-        #         x=df['Model'],
-        #         y=df[col],
-        #         orientation='v',
-        #         marker_color=px.colors.qualitative.Plotly,
-        #         text=df[col].round(1),
-        #         textposition='outside',
-        #     ))
-        # fig.update_layout(
-        #     title="Model Performance - Per Embodiment",
-        #     xaxis_title="Model",
-        #     yaxis_title="Score",
-        #     yaxis_range=[0, 100],
-        #     barmode='group',
-        #     height=500,
-        # )
+
+        # Format df
+        df_fig = df.copy()
+        df_fig = df_fig[df_fig["score"] != np.inf]
+
+        # Calculate the model order
+        df_model_order = df_fig.groupby("model")[["score"]].mean().reset_index()
+        model_order = df_model_order.sort_values(by="score", ascending=True)["models"].tolist()
+
+        # Calculate mean score per model and embodiment
+        df_fig = df_fig.groupby(["model", "embodiment"])[["score"]].mean().reset_index()
+
+        # Sort the results from best to worst
+        df_fig = df_fig.sort_values(by="score", ascending=True)
+
+        # Convert the "model" column to a categorical type with the sorted order
+        df_fig["model"] = pd.Categorical(df_fig["model"], categories=model_order, ordered=True)
+
+        # Sort the DataFrame based on the new categorical order
+        df_fig = df_fig.sort_values(by=["model", "score"], ascending=[True, True])
+
+        # Create the Plotly figure
+        fig = px.bar(
+            df_fig,
+            x="model",
+            y="score",
+            color="embodiment",
+            color_continuous_scale=px.colors.qualitative.Plotly,
+            orientation="v",
+        )
+        max_score = df_fig["score"].max()
+        fig.update_layout(
+            xaxis_title_text="Model",
+            yaxis_title_text="Score (Lower is better)",
+            title_text="",
+            font=dict(size=15),
+            xaxis_tickangle=-45,
+            bargap=0.1,
+            barmode="group",
+            height=600,
+            margin=dict(
+                l=80,   # Left
+                r=0,    # Right
+                b=80,   # Bottom
+                t=70,   # Top
+                pad=0   # Padding
+            ),
+            showlegend=True,
+            legend=dict(
+                x=0,
+                y=1.3,
+                xanchor="left",
+                yanchor="top",
+                bgcolor="rgba(255,255,255,0)",
+                bordercolor="rgba(0,0,0,0)",
+                borderwidth=1,
+                itemclick="toggle",
+                itemdoubleclick="toggleothers",
+                title=dict(
+                    text=f"<b>Embodiments</b>",
+                    font=dict(size=12, color="black"),
+                    side="top center",
+                ),
+                font=dict(size=10, color='black')
+            ),
+            uniformtext_minsize=10,
+            uniformtext_mode="show",
+            yaxis_range=[0, max_score * 1.25]
+        )
+        
+        # Remove the color legend from the chart.
+        fig.update_coloraxes(showscale=False)
+        
+        # Add annotations to show the exact score on each bar.
+        fig.update_traces(
+            texttemplate="%{y:.2f}",
+            textposition="outside",
+            textangle=-90,
+        )
     
     else:  # Per Category
         # category_cols = [col for col in df.columns if col.startswith('Category-')]