About batch_size during training

[zhangziwenHIT]

You mentioned in the paper that setting batch_size==8 on four 2080Ti takes about three days to train. But when I set batch_size>=4 on V100, I get an error. The problem is the same as in this issue.https://github.com/bes-dev/MobileStyleGAN.pytorch/issues/23 Is there something wrong with my configuration?As far as I know, the 2080Ti has 11GB of memory, while the V100 I used has 32GB. This is puzzling to me.Looking forward to getting your answer, thank you!