Update handler.py

handler.py

@@ -23,11 +23,13 @@ class EndpointHandler:
         # preprocess
         input_ids = self.tokenizer(inputs, return_tensors="pt").input_ids
 
+        # postprocess the prediction
+        prediction = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
         # pass inputs with all kwargs in data
         if parameters is not None:
-            prediction = self.pipeline(inputs, device=0, **parameters)
+            outputs = self.model.generate(inputs, device=0, **parameters)
         else:
-            prediction = self.pipeline(inputs, device=0)
+            outputs = self.model.generate(inputs, device=0)
 
         # postprocess the prediction
         prediction = self.tokenizer.decode(outputs[0], skip_special_tokens=True)