I have solved the problem. I have accidentally add “num_words” in the Tokenizer that caused the issue. Sorry for the inconvenience.