Update handler.py

handler.py

@@ -9,7 +9,8 @@ class EndpointHandler():
         self.model = AutoModelForSeq2SeqLM.from_pretrained(path)
 
     def __call__(self, data: str) -> str:
-        inp = self.tokenizer(data, return_tensors="pt")
+        inputs = data.pop("inputs", data)
+        inp = self.tokenizer(inputs, return_tensors="pt")
         with torch.inference_mode():
             out= self.model.generate(**inp)
         final_output = self.tokenizer.batch_decode(out,skip_special_tokens=True)