chore: remove example

app.py

@@ -2,9 +2,7 @@ import spaces
 import torch
 import gradio as gr
 import requests
-from PIL import Image
 from torchvision import transforms, models
-import os
 
 # Download human-readable labels for ImageNet.
 response = requests.get("https://git.io/JJkYN")
@@ -23,51 +21,44 @@ preprocess = transforms.Compose([
 ])
 
 @spaces.GPU(duration=60)
-def gpu_inference(input_tensor):
-    """Isolated GPU function"""
-    device = torch.device("cuda")
-    model.to(device)
-    input_tensor = input_tensor.to(device)
-    
+def run_model_on_gpu(input_tensor):
+    """Pure GPU computation function - no Gradio context"""
     with torch.no_grad():
+        # Move everything to GPU
+        device = torch.device("cuda")
+        model.to(device)
+        input_tensor = input_tensor.to(device)
+        
+        # Run inference
         output = model(input_tensor)
-        prediction = torch.nn.functional.softmax(output[0], dim=0)
-    
-    return prediction.cpu()
+        
+        # Return CPU tensor
+        return output.cpu()
 
 def predict(inp):
-    """Main prediction function"""
+    """Main prediction function that handles all logic"""
     if inp is None:
         return {}
     
-    try:
-        # Preprocess on CPU
-        input_tensor = preprocess(inp).unsqueeze(0)
-        
-        # Run GPU inference
-        prediction = gpu_inference(input_tensor)
-        
-        # Convert to confidences
-        confidences = {labels[i]: float(prediction[i]) for i in range(1000)}
-        return confidences
-    except Exception as e:
-        print(f"Error in prediction: {e}")
-        return {"error": str(e)}
-
-# Create example list only if example files exist
-example_files = ["lion.jpg", "cheetah.jpg", "cat.avif", "hot-dog.avif", "llama.jpg", "medieval_knight.jpg"]
-examples = [f for f in example_files if os.path.exists(f)]
+    # Preprocess image on CPU
+    input_tensor = preprocess(inp).unsqueeze(0)
+    
+    # Get model output via GPU function
+    output = run_model_on_gpu(input_tensor)
+    
+    # Process predictions on CPU
+    prediction = torch.nn.functional.softmax(output[0], dim=0)
+    confidences = {labels[i]: float(prediction[i]) for i in range(1000)}
+    
+    # Return top predictions
+    return confidences
 
-# Create Gradio interface
-iface = gr.Interface(
+# Create Gradio interface without examples first to test
+gr.Interface(
     fn=predict, 
     inputs=gr.Image(type="pil"),
     outputs=gr.Label(num_top_classes=3),
-    examples=examples if examples else None,
-    cache_examples=False,
     title="Image Classifier with ZeroGPU",
+    description="Upload an image to classify it using ResNet-34",
     css=".footer{display:none !important}"
-)
-
-if __name__ == "__main__":
-    iface.launch()
+).launch()