pop image

handler.py

@@ -20,9 +20,9 @@ class EndpointHandler():
       Return:
             A :obj:`list` | `dict`: will be serialized and returned
         """
-
-        words = data.pop("words", data)
-        imageData = data.pop("image", data)
+        inputs = data.pop("inputs", data)
+        words = inputs["words"]
+        imageData = inputs["image"]
         image = Image.open(BytesIO(base64.b64decode(imageData)))
         inputs = self.processor(text=words, images=image, return_tensors="pt", padding=True)
         outputs = self.model(**inputs)