update handler for testing

handler.py

@@ -45,4 +45,5 @@ class EndpointHandler():
         with torch.cuda.amp.autocast():
             # prediction = self.pipeline(inputs, generate_kwargs={"forced_decoder_ids": self.forced_decoder_ids}, max_new_tokens=255)["text"]
             prediction = self.pipeline(inputs, return_timestamps=False)
-        return [prediction, {"test": 0}]
+            prediction['text'] = prediction['text'] + '????'
+        return prediction