Fix inputs

handler.py

@@ -13,8 +13,9 @@ class EndpointHandler:
 
     def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
         # process input
-        image = data.pop("image", data)
-        text = data.pop("text", data)
+        inputs = data.pop("inputs", data)
+        image = inputs["image"]
+        text = inputs["text"]
 
         # preprocess
         encoding = self.processor(image, text, return_tensors="pt")