Remove logic for 'All' and 'Aggregated' for Graph view.

Aggregated scores are computed by each task. In the graph view we do not give the ability to select tasks but to select benchmarks.

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 scale-hf-logo.png filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text

app.py

@@ -76,64 +76,18 @@ def update_benchmarks_by_task(task):
     return gr.update(value=benchmark_value, choices=new_benchmarks), filtered
 
 def generate_scatter_plot(benchmark, metric):
-    if benchmark == "All":
-        models_data = []
-        
-        for bench in benchmarks:
-            subset = df[df['Benchmark'] == bench]
-            if bench == "RTL-Repo":
-                subset = subset[subset['Metric'].str.contains('EM', case=False, na=False)]
-            models_in_bench = subset['Model'].unique()
-            models_data.extend([(model, bench) for model in models_in_bench])
-        
-        all_models = list(set([m[0] for m in models_data]))
-        details = df[['Model', 'Params', 'Model Type']].drop_duplicates('Model')
-        
-        if metric == "Aggregated ⬆️":
-            agg_columns = [col for col in df_agg.columns if col.startswith('Agg ')]
-            if agg_columns:
-                agg_data = df_agg.copy()
-                agg_data['Aggregated ⬆️'] = agg_data[agg_columns].mean(axis=1).round(2)
-                scatter_data = pd.merge(details, agg_data[['Model', 'Aggregated ⬆️']], on='Model', how='inner')
-            else:
-                scatter_data = details.copy()
-                scatter_data['Aggregated ⬆️'] = 50  # defaut
-        else:
-            scatter_data = details.copy()
-            metric_data = df[df['Metric'] == metric].groupby('Model')['Score'].mean().reset_index()
-            metric_data = metric_data.rename(columns={'Score': metric})
-            scatter_data = pd.merge(scatter_data, metric_data, on='Model', how='left')
-        scatter_data = scatter_data.dropna(subset=[metric] if metric in scatter_data.columns else ['Aggregated ⬆️'])
-        
+    benchmark, metric = handle_special_cases(benchmark, metric)
+    
+    subset = df[df['Benchmark'] == benchmark]
+    if benchmark == "RTL-Repo":
+        subset = subset[subset['Metric'].str.contains('EM', case=False, na=False)]
+        detailed_scores = subset.groupby('Model', as_index=False)['Score'].mean()
+        detailed_scores.rename(columns={'Score': 'Exact Matching (EM)'}, inplace=True)
     else:
-        # Code we already had for individual benchmark selection
-        benchmark, metric = handle_special_cases(benchmark, metric)
+        detailed_scores = subset.pivot_table(index='Model', columns='Metric', values='Score').reset_index()
         
-        subset = df[df['Benchmark'] == benchmark]
-        if benchmark == "RTL-Repo":
-            subset = subset[subset['Metric'].str.contains('EM', case=False, na=False)]
-            detailed_scores = subset.groupby('Model', as_index=False)['Score'].mean()
-            detailed_scores.rename(columns={'Score': 'Exact Matching (EM)'}, inplace=True)
-            detailed_scores['Aggregated ⬆️'] = detailed_scores['Exact Matching (EM)']
-        else:
-            agg_column = None
-            detailed_scores = subset.pivot_table(index='Model', columns='Metric', values='Score').reset_index()
-            if benchmark == 'VerilogEval S2R':
-                agg_column = 'Agg VerilogEval S2R'
-            elif benchmark == 'VerilogEval MC':
-                agg_column = 'Agg VerilogEval MC'
-            elif benchmark == 'RTLLM':
-                agg_column = 'Agg RTLLM'
-            elif benchmark == 'VeriGen':
-                agg_column = 'Agg VeriGen'
-            if agg_column and agg_column in df_agg.columns:
-                agg_data = df_agg[['Model', agg_column]].rename(columns={agg_column: 'Aggregated ⬆️'})
-                detailed_scores = pd.merge(detailed_scores, agg_data, on='Model', how='left')
-            else:
-                detailed_scores['Aggregated ⬆️'] = detailed_scores[['Syntax (STX)', 'Functionality (FNC)', 'Synthesis (SYN)', 'Power', 'Performance', 'Area']].mean(axis=1).round(2)
-
-        details = df[['Model', 'Params', 'Model Type']].drop_duplicates('Model')
-        scatter_data = pd.merge(detailed_scores, details, on='Model', how='left').dropna(subset=['Params', metric])
+    details = df[['Model', 'Params', 'Model Type']].drop_duplicates('Model')
+    scatter_data = pd.merge(detailed_scores, details, on='Model', how='left').dropna(subset=['Params', metric])
 
     scatter_data['x'] = scatter_data['Params']
     scatter_data['y'] = scatter_data[metric]
@@ -144,8 +98,7 @@ def generate_scatter_plot(benchmark, metric):
 
     y_axis_limits = {
         'Functionality (FNC)': [5, 90], 'Syntax (STX)': [20, 100], 'Synthesis (SYN)': [5, 90],
-        'Power': [0, 50], 'Performance': [0, 50], 'Area': [0, 50], 'Exact Matching (EM)': [0, 50],
-        'Aggregated ⬆️': [0, 80]
+        'Power': [0, 50], 'Performance': [0, 50], 'Area': [0, 50], 'Exact Matching (EM)': [0, 50]
     }
     y_range = y_axis_limits.get(metric, [0, 80])
 
@@ -190,8 +143,8 @@ with gr.Blocks(css=custom_css, js=js_func, theme=gr.themes.Default(primary_hue=c
     s2r_benchs = ["VerilogEval S2R", "RTLLM"]
     cc_benchs = ["VerilogEval MC", "VeriGen"]
     lc_benchs = ["RTL-Repo"]
+    non_rtl_metrics = ["Syntax (STX)", "Functionality (FNC)", "Synthesis (SYN)", "Power", "Performance", "Area"]
     rtl_metrics = ["Exact Matching (EM)"]
-    non_rtl_metrics = ["Syntax (STX)", "Functionality (FNC)", "Synthesis (SYN)", "Power", "Performance", "Area", "Aggregated ⬆️"]
     model_types = ['All', 'General', 'Coding', 'RTL-Specific']
     
     gr.HTML("""
@@ -263,10 +216,12 @@ with gr.Blocks(css=custom_css, js=js_func, theme=gr.themes.Default(primary_hue=c
         
         with gr.Tab("Interactive Bubble Plot"):
             with gr.Row(equal_height=True):
-                bubble_benchmark = gr.Dropdown(choices=["All"] + benchmarks, label="Select Benchmark", value='All', elem_classes="gr-dropdown")
-                bubble_metric = gr.Dropdown(choices=["Aggregated ⬆️"] + non_rtl_metrics[:-1], label="Select Metric", value="Aggregated ⬆️")
+                default_benchmark = s2r_benchs[0]
+                bubble_benchmark = gr.Dropdown(choices=benchmarks, label="Select Benchmark", value=default_benchmark, elem_classes="gr-dropdown")
+                default_metric = non_rtl_metrics[0]
+                bubble_metric = gr.Dropdown(choices=non_rtl_metrics[:-1], label="Select Metric", value=default_metric)
             with gr.Row(equal_height=True):
-                scatter_plot = gr.Plot(value=generate_scatter_plot('All', "Aggregated ⬆️"), label="Bubble Chart", elem_id="full-width-plot")
+                scatter_plot = gr.Plot(value=generate_scatter_plot(default_benchmark, default_metric), label="Bubble Chart", elem_id="full-width-plot")
 
         with gr.Tab("About Us"):
             gr.HTML(
@@ -315,18 +270,13 @@ with gr.Blocks(css=custom_css, js=js_func, theme=gr.themes.Default(primary_hue=c
     search_box.change(fn=filter_leaderboard, inputs=[task_radio, benchmark_radio, model_type_dropdown, search_box, params_slider], outputs=leaderboard)
     params_slider.change(fn=filter_leaderboard, inputs=[task_radio, benchmark_radio, model_type_dropdown, search_box, params_slider], outputs=leaderboard)
 
-    # RTL-Repo Bubble plot
     def on_benchmark_change(benchmark, _):
         if benchmark == "RTL-Repo":
             metric = "Exact Matching (EM)"
             return gr.update(choices=rtl_metrics, value=metric), generate_scatter_plot(benchmark, metric)
         else:
-            if benchmark == "All":
-                metric = "Aggregated ⬆️" # default to Aggregated
-                return gr.update(choices=["Aggregated ⬆️"] + non_rtl_metrics[:-1], value=metric), generate_scatter_plot(benchmark, metric)
-            else:
-                metric = non_rtl_metrics[0]
-                return gr.update(choices=non_rtl_metrics[:-1], value=metric), generate_scatter_plot(benchmark, metric)
+            metric = non_rtl_metrics[0]
+            return gr.update(choices=non_rtl_metrics[:-1], value=metric), generate_scatter_plot(benchmark, metric)
 
     def on_metric_change(benchmark, metric):
         benchmark, metric = handle_special_cases(benchmark, metric)