The first token missed while testing

[zhuochunli]

Hi, why was the first token missed in my inference output? I used your ways to preprocess all my own data and trained it based on your command. But when I tested it, the output always showed that the very first token was missed in the final output. But the output seems correct when using your provided trained model to infer. So I wonder what could I do to solve this? Thank you! E.g. is a man who comes to the park.