Update handler.py

handler.py

@@ -19,10 +19,16 @@ class EndpointHandler():
         self.tokenizer = AutoTokenizer.from_pretrained(cfg['repo'])
 
     def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
-        inputs = data.pop("inputs", "Q: What is the chemical composition of common concrete in 2024?\nA: ")
+        """
+       data args:
+            inputs (:obj: `str` | `PIL.Image` | `np.array`)
+            kwargs
+      Return:
+            A :obj:`list` | `dict`: will be serialized and returned
+        """
         max_new_tokens = 1024
         if "max_new_tokens" in data:
-            max_new_tokens = data.pop("max_new_tokens", 0)
+            max_new_tokens = data["max_new_tokens"]
         try:
             max_new_tokens = int(max_new_tokens)
         except Exception as e: