Updated lines 27 , 37, and 38 with: error handling -> streamlining meaning feedback if an error is logged obverse model fomenting Pokémon predictions.

app.py

@@ -24,15 +24,19 @@ def preprocess_image(img_pil):
 
 # Function to predict the class of the image
 def predict_classification(img_pil):
-    inputs = preprocess_image(img_pil)
-    pixel_values = inputs['pixel_values'][0]  # Extract the pixel_values tensor
-    with torch.no_grad():
-        outputs = model(pixel_values=pixel_values.unsqueeze(0))  # Add batch dimension
-    logits = outputs.logits
-    predicted_class_idx = logits.argmax(-1).item()
-    predicted_class = labels[str(predicted_class_idx)]
-    confidence = torch.nn.functional.softmax(logits, dim=1).numpy()[0][predicted_class_idx]
-    return predicted_class, confidence
+    try:
+        inputs = preprocess_image(img_pil)
+        pixel_values = inputs['pixel_values'][0]  # Extract the pixel_values tensor
+        with torch.no_grad():
+            outputs = model(pixel_values=pixel_values.unsqueeze(0))  # Add batch dimension
+        logits = outputs.logits
+        predicted_class_idx = logits.argmax(-1).item()
+        predicted_class = labels[str(predicted_class_idx)]
+        confidence = torch.nn.functional.softmax(logits, dim=1).numpy()[0][predicted_class_idx]
+        return predicted_class, confidence
+    except Exception as e: 
+        return "Error occurred during prediction", 0.0 
+    
 
 # Function to handle the prediction in the Gradio interface
 def pokémon_predict(img_pil):