ml_predict in BCILAB

In the ml_predict script of BCILAB , when there is no label as input argument , how am I getting the prediction accuracy ? How is the script calculating it without the user giving it the labels . someone please shed some light on it . Thank you.