"inputs" is required

handler.py

@@ -14,8 +14,9 @@ class EndpointHandler():
         self.processor = CLIPProcessor.from_pretrained("rbanfield/clip-vit-large-patch14")
 
     def __call__(self, data):
-        text_input = data.pop("text", None)
-        image_input = data.pop("image", None)
+        inputs = data.pop("inputs", None)
+        text_input = inputs["text"] if "text" in inputs else None
+        image_input = inputs["image"] if "image" in inputs else None
 
         if text_input:
             processor = self.processor(text=text_input, return_tensors="pt", padding=True)
@@ -28,3 +29,4 @@ class EndpointHandler():
                 return self.image_model(**processor).image_embeds.tolist()
         else:
             return None
+[rbanfield@atlas clip-vit-large-pa