Update app.py

app.py

@@ -80,7 +80,6 @@ import gradio as gr
 # Authentication token for HuggingFace API access
 # Set via environment variable: export HF_TOKEN="your_token_here"
 HF_TOKEN = os.getenv("HF_TOKEN", "")
-
 # =================================================================================================
 # UI STYLING
 # =================================================================================================
@@ -1565,14 +1564,14 @@ def process_request(
     output_tokens: float,
     dataset_tier: str,
     manufacturers: list[str],
-) -> Tuple[str, str, str, Any, Any, Any]: #, Any]:
+) -> Tuple[str, str, str, Any]:
     """
     Main request processing function.
     
     Orchestrates model resolution, VRAM calculation, and hardware recommendation.
     
     Returns:
-        Tuple of (report, budget_rec, runner_up_rec, throughput_chart, cost_chart, vram_chart, efficiency_chart)
+        Tuple of (report, budget_rec, runner_up_rec, combined_chart)
     """
     try:
         # Validate inputs
@@ -1630,7 +1629,7 @@ def process_request(
         )
         
         if error:
-            return error, "", "", None, None, None, None
+            return error, "", "", None
         
         # Generate report
         report = f"""
@@ -1661,84 +1660,63 @@ def process_request(
 *ℹ️ NOTE: VRAM estimates include 10% safety buffer. Actual usage may vary ±15-20% based on framework optimizations, model architecture details, and runtime conditions. Throughput estimates are based on empirical benchmarks and may vary in production.*
         """
         
-        # Create bar charts using Plotly
-        throughput_chart = None
-        cost_chart = None
-        vram_chart = None
-        efficiency_chart = None
+        # Create grouped multi-bar chart using Plotly
+        combined_chart = None
         
         if chart_data and chart_data.get("names"):
             import plotly.graph_objects as go
             
-            # Throughput chart
-            throughput_chart = go.Figure(data=[
+            # Normalize values to percentages for comparison (0-100 scale)
+            max_throughput = max(chart_data["throughput"]) if chart_data["throughput"] else 1
+            max_cost = max(chart_data["cost"]) if chart_data["cost"] else 1
+            
+            throughput_normalized = [(t / max_throughput) * 100 for t in chart_data["throughput"]]
+            cost_normalized = [(c / max_cost) * 100 for c in chart_data["cost"]]
+            
+            combined_chart = go.Figure(data=[
                 go.Bar(
+                    name='Throughput',
                     x=chart_data["names"],
-                    y=chart_data["throughput"],
-                    marker_color='#22c55e'
-                )
-            ])
-            throughput_chart.update_layout(
-                title="Throughput Comparison",
-                xaxis_title="GPU Configuration",
-                yaxis_title="Tokens/sec",
-                height=350,
-                xaxis_tickangle=-45,
-                margin=dict(b=120)
-            )
-            
-            # Cost chart
-            cost_chart = go.Figure(data=[
+                    y=throughput_normalized,
+                    marker_color='#22c55e',
+                    hovertemplate='%{x}<br>Throughput: %{customdata:,.0f} tok/s<extra></extra>',
+                    customdata=chart_data["throughput"]
+                ),
                 go.Bar(
+                    name='Cost',
                     x=chart_data["names"],
-                    y=chart_data["cost"],
-                    marker_color='#3b82f6'
-                )
-            ])
-            cost_chart.update_layout(
-                title="Cost Comparison",
-                xaxis_title="GPU Configuration",
-                yaxis_title="₹/hour",
-                height=350,
-                xaxis_tickangle=-45,
-                margin=dict(b=120)
-            )
-            
-            # VRAM utilization chart
-            vram_chart = go.Figure(data=[
+                    y=cost_normalized,
+                    marker_color='#3b82f6',
+                    hovertemplate='%{x}<br>Cost: ₹%{customdata:,.2f}/hr<extra></extra>',
+                    customdata=chart_data["cost"]
+                ),
                 go.Bar(
+                    name='VRAM Utilization',
                     x=chart_data["names"],
                     y=chart_data["vram_util"],
-                    marker_color='#a855f7'
+                    marker_color='#a855f7',
+                    hovertemplate='%{x}<br>VRAM Util: %{y:.1f}%<extra></extra>'
                 )
             ])
-            vram_chart.update_layout(
-                title="VRAM Utilization Comparison",
+            
+            combined_chart.update_layout(
+                title_text="GPU Comparison (Top 10 by Cost)",
                 xaxis_title="GPU Configuration",
-                yaxis_title="Utilization %",
-                height=350,
+                yaxis_title="Normalized Value (% of max)",
+                barmode='group',
+                height=450,
                 xaxis_tickangle=-45,
-                margin=dict(b=120)
-            )
-            
-            # Cost efficiency chart
-            efficiency_chart = go.Figure(data=[
-                go.Bar(
-                    x=chart_data["names"],
-                    y=chart_data["cost_efficiency"],
-                    marker_color='#f59e0b'
+                margin=dict(b=120),
+                legend=dict(
+                    orientation="h",
+                    yanchor="bottom",
+                    y=1.02,
+                    xanchor="center",
+                    x=0.5
                 )
-            ])
-            # efficiency_chart.update_layout(
-            #     title="Cost Efficiency Comparison",
-            #     xaxis_title="GPU Configuration",
-            #     yaxis_title="Throughput / Cost",
-            #     height=350,
-            #     xaxis_tickangle=-45,
-            #     margin=dict(b=120)
-            # )
+            )
         
-        return report, budget_rec, runner_up_rec, throughput_chart, cost_chart, vram_chart #, efficiency_chart
+        return report, budget_rec, runner_up_rec, combined_chart
         
     except Exception as e:
         error_msg = f"""
@@ -1747,7 +1725,7 @@ def process_request(
 <p>{str(e)}</p>
 </div>
         """
-        return error_msg, "", "", None, None, None, None
+        return error_msg, "", "", None
 
 # =================================================================================================
 # UI Event Handlers
@@ -1909,17 +1887,9 @@ def create_interface() -> gr.Blocks:
                     with gr.Column(elem_classes=["runner-box"]):
                         rec_out_2 = gr.Markdown()
                 
-                # Bar chart comparison section with tabs inside Accordion
+                # Bar chart comparison section
                 with gr.Accordion("📊 GPU Comparison (Top 10 by Cost)", open=False):
-                    with gr.Tabs():
-                        with gr.Tab("Cost"):
-                            cost_plot = gr.Plot(label="Cost Comparison")
-                        with gr.Tab("Throughput"):
-                            throughput_plot = gr.Plot(label="Throughput Comparison")
-                        with gr.Tab("VRAM Utilization"):
-                            vram_plot = gr.Plot(label="VRAM Utilization Comparison")
-                        # with gr.Tab("Cost Efficiency"):
-                        #     efficiency_plot = gr.Plot(label="Cost Efficiency Comparison")
+                    combined_plot = gr.Plot(label="GPU Comparison")
 
                 with gr.Accordion("📈 Details", open=False):
                     report_out = gr.Markdown()
@@ -1955,7 +1925,7 @@ def create_interface() -> gr.Blocks:
                 ft_method, rank, batch, sample_count,
                 input_tokens, output_tokens, dataset_tier, manufacturers,
             ],
-            outputs=[report_out, rec_out_1, rec_out_2, throughput_plot, cost_plot, vram_plot] #, efficiency_plot]
+            outputs=[report_out, rec_out_1, rec_out_2, combined_plot]
         )
     
     return demo