Spaces:
Running
on
Zero
Running
on
Zero
Suggestions Needed.
#4
by
DyrusQZ
- opened
Welcome to leave the comments about the improved points to really adopt LHM for animation video generation~
hi thanks for sharing codes.
I'm not sure inserting print method make my duplicate space stable.
AI said it's need synchronize.however Zero-GPU always make strange error,It could be a coincidence.
rgb = batch_dict["comp_rgb"].detach().cpu().numpy() # [Nv, H, W, 3], 0-1
torch.cuda.synchronize()
#print("rgb")
mask = batch_dict["comp_mask"].detach().cpu().numpy() # [Nv, H, W, 3], 0-1
#print("mask")