Update app.py
Browse files
app.py
CHANGED
|
@@ -119,36 +119,34 @@ def explainability(_):
|
|
| 119 |
shap_values = explainer.shap_values(X_test)
|
| 120 |
|
| 121 |
try:
|
| 122 |
-
if isinstance(shap_values, list):
|
| 123 |
class_idx = 0
|
| 124 |
-
|
| 125 |
-
X_test.values[:, :shap_values[class_idx].shape[1]],
|
| 126 |
-
columns=X_test.columns[:shap_values[class_idx].shape[1]]
|
| 127 |
-
)
|
| 128 |
-
shap.summary_plot(shap_values[class_idx], X_shap, show=False)
|
| 129 |
-
shap_path = f"./shap_class_{class_idx}.png"
|
| 130 |
-
plt.title(f"SHAP Summary - Class {class_idx}")
|
| 131 |
else:
|
| 132 |
-
|
| 133 |
-
X_test.values[:, :shap_values.shape[1]],
|
| 134 |
-
columns=X_test.columns[:shap_values.shape[1]]
|
| 135 |
-
)
|
| 136 |
-
shap.summary_plot(shap_values, X_shap, show=False)
|
| 137 |
-
shap_path = "./shap_plot.png"
|
| 138 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 139 |
plt.savefig(shap_path)
|
| 140 |
-
if wandb.run
|
| 141 |
wandb.log({"shap_summary": wandb.Image(shap_path)})
|
| 142 |
plt.clf()
|
| 143 |
|
| 144 |
except Exception as e:
|
| 145 |
shap_path = "./shap_error.png"
|
| 146 |
-
print(
|
| 147 |
plt.figure(figsize=(6, 3))
|
| 148 |
-
plt.text(0.5, 0.5, f"SHAP
|
| 149 |
plt.axis('off')
|
| 150 |
plt.savefig(shap_path)
|
| 151 |
-
if wandb.run
|
| 152 |
wandb.log({"shap_error": wandb.Image(shap_path)})
|
| 153 |
plt.clf()
|
| 154 |
|
|
@@ -163,7 +161,7 @@ def explainability(_):
|
|
| 163 |
lime_fig = lime_exp.as_pyplot_figure()
|
| 164 |
lime_path = "./lime_plot.png"
|
| 165 |
lime_fig.savefig(lime_path)
|
| 166 |
-
if wandb.run
|
| 167 |
wandb.log({"lime_explanation": wandb.Image(lime_path)})
|
| 168 |
plt.clf()
|
| 169 |
|
|
@@ -171,6 +169,7 @@ def explainability(_):
|
|
| 171 |
|
| 172 |
|
| 173 |
|
|
|
|
| 174 |
with gr.Blocks() as demo:
|
| 175 |
gr.Markdown("## 📊 AI-Powered Data Analysis with Hyperparameter Optimization")
|
| 176 |
|
|
|
|
| 119 |
shap_values = explainer.shap_values(X_test)
|
| 120 |
|
| 121 |
try:
|
| 122 |
+
if isinstance(shap_values, list):
|
| 123 |
class_idx = 0
|
| 124 |
+
sv = shap_values[class_idx]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
else:
|
| 126 |
+
sv = shap_values
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
|
| 128 |
+
# Align number of columns
|
| 129 |
+
if sv.shape[1] != X_test.shape[1]:
|
| 130 |
+
X_test_trimmed = pd.DataFrame(X_test.values[:, :sv.shape[1]], columns=X_test.columns[:sv.shape[1]])
|
| 131 |
+
else:
|
| 132 |
+
X_test_trimmed = X_test
|
| 133 |
+
|
| 134 |
+
shap.summary_plot(sv, X_test_trimmed, show=False)
|
| 135 |
+
shap_path = "./shap_plot.png"
|
| 136 |
+
plt.title("SHAP Summary")
|
| 137 |
plt.savefig(shap_path)
|
| 138 |
+
if wandb.run:
|
| 139 |
wandb.log({"shap_summary": wandb.Image(shap_path)})
|
| 140 |
plt.clf()
|
| 141 |
|
| 142 |
except Exception as e:
|
| 143 |
shap_path = "./shap_error.png"
|
| 144 |
+
print("SHAP plotting failed:", e)
|
| 145 |
plt.figure(figsize=(6, 3))
|
| 146 |
+
plt.text(0.5, 0.5, f"SHAP Error:\n{str(e)}", ha='center', va='center')
|
| 147 |
plt.axis('off')
|
| 148 |
plt.savefig(shap_path)
|
| 149 |
+
if wandb.run:
|
| 150 |
wandb.log({"shap_error": wandb.Image(shap_path)})
|
| 151 |
plt.clf()
|
| 152 |
|
|
|
|
| 161 |
lime_fig = lime_exp.as_pyplot_figure()
|
| 162 |
lime_path = "./lime_plot.png"
|
| 163 |
lime_fig.savefig(lime_path)
|
| 164 |
+
if wandb.run:
|
| 165 |
wandb.log({"lime_explanation": wandb.Image(lime_path)})
|
| 166 |
plt.clf()
|
| 167 |
|
|
|
|
| 169 |
|
| 170 |
|
| 171 |
|
| 172 |
+
|
| 173 |
with gr.Blocks() as demo:
|
| 174 |
gr.Markdown("## 📊 AI-Powered Data Analysis with Hyperparameter Optimization")
|
| 175 |
|