in the PSGAN paper the makeup loss is denoted as using an l2 loss  However, in the implementation an L1 criterion is used. Which loss is correct?
in the PSGAN paper the makeup loss is denoted as using an l2 loss

However, in the implementation an L1 criterion is used.
Which loss is correct?