|
mi_loss = self.bce_loss(logits * mask, labels * mask) * (batch_size * (pos_size + neg_size)) \ |
hi~ Thanks for your clear codes. I have two questions and hope for you to reply.
1.why Line 58 above multiply the (batch_size * (pos_size + neg_size))?? In your paper, this item didn't show up.
2. I saw a lot of mask in your codes, but I'm not much clear about the meaning of mask.
GroupIM/models/discriminator.py
Line 58 in 899890e
hi~ Thanks for your clear codes. I have two questions and hope for you to reply.
1.why Line 58 above multiply the (batch_size * (pos_size + neg_size))?? In your paper, this item didn't show up.
2. I saw a lot of mask in your codes, but I'm not much clear about the meaning of mask.