Update app.py

app.py

@@ -18,7 +18,6 @@ try:
     model.eval()
     model = model.to(device)
     
-    # Use half precision on GPU to save memory
     if device.type == "cuda":
         model = model.half()
         print("Using half precision (FP16) for GPU inference")
@@ -52,7 +51,6 @@ def process_image(input_img):
         # Preprocess image
         image_tensor = preprocess(input_img).unsqueeze(0).to(device)
         
-        # Use appropriate dtype
         if device.type == "cuda":
             image_tensor = image_tensor.half()
         
@@ -64,10 +62,7 @@ def process_image(input_img):
             else:
                 image_features = model.encode_image(image_tensor)
             
-            # Normalize features
             image_features = image_features / image_features.norm(dim=-1, keepdim=True)
-            
-            # Convert to numpy
             embedding = image_features.cpu().float().numpy().flatten()
         
         # Clean up GPU memory
@@ -75,47 +70,42 @@ def process_image(input_img):
         if torch.cuda.is_available():
             torch.cuda.empty_cache()
         
-        # Create simple result string instead of complex JSON
-        result_text = f"""✅ Embedding Extracted Successfully!
-
-📊 Results:
-- Model: MobileCLIP-S2
-- Device: {device}
-- Embedding Dimension: {len(embedding)}
-- Mean: {np.mean(embedding):.4f}
-- Std: {np.std(embedding):.4f}
-- Min: {np.min(embedding):.4f}
-- Max: {np.max(embedding):.4f}
+        # Simple result format that works with older Gradio
+        result_text = f"""Embedding Extracted Successfully!
 
-🔢 Preview (first 10 values):
-{embedding[:10].tolist()}
+Model: MobileCLIP-S2
+Device: {device}
+Embedding Dimension: {len(embedding)}
+Mean: {np.mean(embedding):.4f}
+Std: {np.std(embedding):.4f}
+Min: {np.min(embedding):.4f}
+Max: {np.max(embedding):.4f}
 
-💡 Usage: The full {len(embedding)}-dimensional embedding vector can be used for similarity search, classification, or clustering tasks."""
+First 10 values:
+{embedding[:10].tolist()}"""
         
         return result_text
         
     except Exception as e:
         if torch.cuda.is_available():
             torch.cuda.empty_cache()
-        return f"❌ Error processing image: {str(e)}"
-
-# Create simple Blocks interface
-demo = gr.Blocks(title="MobileCLIP Embeddings")
+        return f"Error processing image: {str(e)}"
 
-with demo:
-    gr.Markdown("# 🚀 MobileCLIP Image Embedding Extractor")
-    gr.Markdown("Extract 512-dimensional embeddings from images using MobileCLIP-S2")
-    
-    with gr.Row():
-        image_input = gr.Image(type="pil", label="Upload Image")
-        text_output = gr.Textbox(label="Results", lines=12, interactive=False)
-    
-    btn = gr.Button("Extract Embeddings")
-    btn.click(process_image, inputs=image_input, outputs=text_output)
+# Use gr.Interface instead of gr.Blocks for better compatibility with 4.16.0
+demo = gr.Interface(
+    fn=process_image,
+    inputs=gr.Image(type="pil", label="Upload Image"),
+    outputs=gr.Textbox(label="Results", lines=15),
+    title="MobileCLIP Image Embedding Extractor",
+    description="Extract 512-dimensional embeddings from images using MobileCLIP-S2",
+    allow_flagging="never"  # Disable flagging to avoid potential issues
+)
 
 if __name__ == "__main__":
     demo.launch(
         server_name="0.0.0.0",
         server_port=7860,
-        share=True
-    )
+        share=False,  # Avoid the HF Spaces warning
+        inbrowser=False,
+        quiet=False
+    )