Update app.py

app.py

@@ -102,7 +102,7 @@ def train_model(_):
 def explainability(_):
     import warnings
     warnings.filterwarnings("ignore")
-    
+
     target = df_global.columns[-1]
     X = df_global.drop(target, axis=1)
     y = df_global[target]
@@ -115,35 +115,44 @@ def explainability(_):
     model = RandomForestClassifier()
     model.fit(X_train, y_train)
 
-    # SHAP Explainability
     explainer = shap.TreeExplainer(model)
     shap_values = explainer.shap_values(X_test)
 
     try:
         if isinstance(shap_values, list):  # Multiclass
             class_idx = 0
-            shap_matrix = shap_values[class_idx]  # shape: (samples, features)
-            assert shap_matrix.shape[1] == X_test.shape[1], \
-                f"SHAP mismatch: shap {shap_matrix.shape} vs X {X_test.shape}"
-            shap.summary_plot(shap_matrix, X_test, show=False)
+            X_shap = pd.DataFrame(
+                X_test.values[:, :shap_values[class_idx].shape[1]],
+                columns=X_test.columns[:shap_values[class_idx].shape[1]]
+            )
+            shap.summary_plot(shap_values[class_idx], X_shap, show=False)
             shap_path = f"./shap_class_{class_idx}.png"
             plt.title(f"SHAP Summary - Class {class_idx}")
         else:
-            shap.summary_plot(shap_values, X_test, show=False)
+            X_shap = pd.DataFrame(
+                X_test.values[:, :shap_values.shape[1]],
+                columns=X_test.columns[:shap_values.shape[1]]
+            )
+            shap.summary_plot(shap_values, X_shap, show=False)
             shap_path = "./shap_plot.png"
+
         plt.savefig(shap_path)
-        wandb.log({"shap_summary": wandb.Image(shap_path)})
+        if wandb.run is not None:
+            wandb.log({"shap_summary": wandb.Image(shap_path)})
         plt.clf()
+
     except Exception as e:
         shap_path = "./shap_error.png"
         print(f"SHAP plotting failed: {e}")
         plt.figure(figsize=(6, 3))
-        plt.text(0.5, 0.5, f"SHAP error:\n{str(e)}", ha='center', va='center')
+        plt.text(0.5, 0.5, f"SHAP error:\\n{str(e)}", ha='center', va='center')
+        plt.axis('off')
         plt.savefig(shap_path)
-        wandb.log({"shap_error": wandb.Image(shap_path)})
+        if wandb.run is not None:
+            wandb.log({"shap_error": wandb.Image(shap_path)})
         plt.clf()
 
-    # LIME Explainability
+    # LIME
     lime_explainer = lime.lime_tabular.LimeTabularExplainer(
         X_train.values,
         feature_names=X_train.columns.tolist(),
@@ -154,12 +163,14 @@ def explainability(_):
     lime_fig = lime_exp.as_pyplot_figure()
     lime_path = "./lime_plot.png"
     lime_fig.savefig(lime_path)
-    wandb.log({"lime_explanation": wandb.Image(lime_path)})
+    if wandb.run is not None:
+        wandb.log({"lime_explanation": wandb.Image(lime_path)})
     plt.clf()
 
     return shap_path, lime_path
 
 
+
 with gr.Blocks() as demo:
     gr.Markdown("## 📊 AI-Powered Data Analysis with Hyperparameter Optimization")