question regarding `assert gold_status is not None`

https://github.com/donote/pytorch-conditional-model/blob/87c00abb022787cfe398ebd5f48e218a3abfcbb1/model.py#L100
Could you explain why do you need this assert, only for the case of training mode? 
The following line is for both cases.
        gold_status = gold_status.view(-1)