Update app.py

app.py

@@ -56,7 +56,9 @@ def update_explorer(dataset_name, split_name):
     splits = df["split"].unique().tolist() if "split" in df.columns else ["all"]
     
     filtered = df[df["split"] == split_name].head(10) if "split" in df.columns else df.head(10)
-    qasm_sample = filtered["circuit_qasm"].iloc[0] if "circuit_qasm" in filtered.columns else "// QASM not found"
+    
+    qasm_col = "circuit_qasm" if "circuit_qasm" in df.columns else None
+    qasm_sample = filtered[qasm_col].iloc[0] if qasm_col and not filtered.empty else "// QASM not available in this pack"
     
     return gr.update(choices=splits), filtered, qasm_sample
 
@@ -84,6 +86,7 @@ def run_model_demo(dataset_name):
     importances = model.feature_importances_
     indices = np.argsort(importances)[-10:]
     ax2.barh(range(10), importances[indices], color='#EF553B')
+    ax2.set_yticks(range(10))
     ax2.set_yticklabels([feature_cols[i] for i in indices])
     ax2.set_title("Top 10 Features")
 
@@ -96,7 +99,7 @@ def run_model_demo(dataset_name):
 
 def load_benchmark():
     path = Path(LOCAL_BENCHMARK_CSV)
-    if not path.exists(): return None, None, None
+    if not path.exists(): return None, None, "File noise_benchmark_results.csv not found."
     df = pd.read_csv(path)
     fig_r2, ax = plt.subplots(figsize=(8, 4))
     ax.bar(df["dataset"], df["r2"], color=['#636EFA', '#EF553B', '#00CC96', '#AB63FA'])
@@ -107,7 +110,7 @@ def load_benchmark():
 # =========================================================
 # INTERFACE
 # =========================================================
-with gr.Blocks(title="QSBench Unified Explorer", theme=gr.themes.Soft()) as demo:
+with gr.Blocks(title="QSBench Unified Explorer") as demo:
     gr.Markdown("# 🌌 QSBench: Quantum Synthetic Benchmark Explorer\n**Professional-grade datasets for Noise-Aware QML and Hardware Optimization.**")
 
     with gr.Tabs():
@@ -117,12 +120,13 @@ with gr.Blocks(title="QSBench Unified Explorer", theme=gr.themes.Soft()) as demo
                 split_selector = gr.Dropdown(choices=["train", "test", "validation"], value="train", label="Split")
             
             data_table = gr.Dataframe(interactive=False)
-            qasm_view = gr.Code(label="Circuit QASM Preview (First row of selection)", language="wasm")
+            qasm_view = gr.Code(label="Circuit QASM Preview (First row of selection)", language="python")
             
             ds_selector.change(update_explorer, [ds_selector, split_selector], [split_selector, data_table, qasm_view])
             split_selector.change(update_explorer, [ds_selector, split_selector], [split_selector, data_table, qasm_view])
 
         with gr.TabItem("🤖 ML Baseline Demo"):
+            gr.Markdown("Train a Random Forest regressor to evaluate how well structural circuit features predict expectation values.")
             model_ds_selector = gr.Dropdown(choices=list(DATASET_MAP.keys()), value="Core (Clean)", label="Select Target Pack")
             train_btn = gr.Button("Train Baseline Model", variant="primary")
             plot_output = gr.Plot()
@@ -130,6 +134,7 @@ with gr.Blocks(title="QSBench Unified Explorer", theme=gr.themes.Soft()) as demo
             train_btn.click(run_model_demo, [model_ds_selector], [plot_output, text_output])
 
         with gr.TabItem("📊 Cross-Dataset Benchmark"):
+            gr.Markdown("Comparison of model performance across different noise environments and hardware transpilation stages.")
             bench_btn = gr.Button("Analyze Robustness Across All Packs")
             bench_table = gr.Dataframe()
             bench_plot = gr.Plot()
@@ -137,19 +142,15 @@ with gr.Blocks(title="QSBench Unified Explorer", theme=gr.themes.Soft()) as demo
 
     gr.Markdown("""
     ---
-    ### 🔬 Data Integrity & Research Value
-    The demo files provided here serve as a **structural validation** for researchers. 
-    - **Demographic**: 8-10 Qubits, Depth 6-8.
-    - **Features**: Includes gate entropy, Meyer-Wallach entanglement, and transpilation metrics.
+    ### 🔬 Research Resources
+    This interface provides a structural overview of the QSBench dataset family. These datasets are designed to support reproducible research in quantum error mitigation and machine learning.
     
-    To achieve state-of-the-art results in error mitigation or noise modeling, access to the full dataset family (up to 200,000 samples) is recommended to ensure statistical significance and model generalization.
-        
-        🔗 [Website](https://qsbench.github.io) | 🤗 [Hugging Face](https://huggingface.co/QSBench) | 🛠️ [GitHub](https://github.com/QSBench/QSBench-Demo)
-        """
-    )
+    - **GitHub**: [QSBench/QSBench-Demo](https://github.com/QSBench/QSBench-Demo)
+    - **Website**: [qsbench.github.io](https://qsbench.github.io)
+    - **Hugging Face**: [Explore all datasets](https://huggingface.co/QSBench)
+    """)
 
-    # Initial load
     demo.load(update_explorer, [ds_selector, split_selector], [split_selector, data_table, qasm_view])
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(theme=gr.themes.Soft())