Yi
Yi
https://github.com/primepake/wav2lip_288x288/blob/12b90f2b77a2aa0fa4019bf96e796fdac3ad0f95/hq_wav2lip_sam_train.py#L310C43-L310C43 Line 310: running_sync_loss, running_l1_loss, running_perceptual_loss = 0., 0., 0., 0. Hi, is this hq_wav2lip_sam_train.py premature to use? should I try the "not sam" hq version first? -oyiyi
Thank you for the wonderful work! Do you happen to have streaming test code available (or design / workflow) for this work? Thanks!
Hi curious if anyone can share your best loss rates across 3 (4) trainings and how you get there?