r=0?

[cxw-droid]

Hi, thank you for your paper and sharing the code. May I ask some questions?

1. In the code below
   

   ect/training/loss.py

   Line 56 in cb35360

   r = self.t_to_r(t)

   
   Is it possible for r to be exactly 0? it seems it is approaching 0, but not exactly 0. If it cannot get value 0, how the consistency model knows the ground truth x_0? I am not sure if there is some misunderstanding here.

2. In the original consistency models paper, the minimum t is 0.002 and in this paper it can be 0. Would you like to expand more on the difference here? Thanks.