You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
感谢好文,我在想,以256x256为例,从[1,2,.....,16]的codebook size,在计算的时候也是resize到16计算残差,这种resize之后的信息,是不是也可以等价于latent diffusion?从大小为1的最粗糙的信息开始到16,这不就是一个类似diffusion的coarse to fine的过程吗?
The text was updated successfully, but these errors were encountered:
感谢好文,我在想,以256x256为例,从[1,2,.....,16]的codebook size,在计算的时候也是resize到16计算残差,这种resize之后的信息,是不是也可以等价于latent diffusion?从大小为1的最粗糙的信息开始到16,这不就是一个类似diffusion的coarse to fine的过程吗?
The text was updated successfully, but these errors were encountered: