Fixed everything.

app.py

@@ -182,8 +182,9 @@ def stream_chat(input_image: Image.Image, caption_type: str, caption_length: str
 
     # Preprocess image
     image = input_image.resize((384, 384), Image.LANCZOS)
+    image = image.convert('RGB')  # Ensure the image has 3 channels
     pixel_values = TVF.pil_to_tensor(image).unsqueeze(0) / 255.0
-    pixel_values = TVF.normalize(pixel_values, [0.5], [0.5])
+    pixel_values = TVF.normalize(pixel_values, [0.5, 0.5, 0.5], [0.5, 0.5, 0.5])  # Normalize for all 3 channels
     pixel_values = pixel_values.to('cuda', dtype=torch.float16)
 
     # Embed image