about the formula

[seaman1900]

RectifiedFlow/ImageGeneration/losses.py

Line 101 in 9df93aa

t = torch.rand(batch.shape[0], device=batch.device) * (sde.T - eps) + eps

You mentioned that xt = (1-t)x + ty, and in the code, t starts with eps (1e-3) in training, however, you used t=eps in inference and the input of model should be x_0 * (1-eps)+ y*eps accordingly. I feel quiet strange about this.

在采样的时候用的是xt = (1-t)x + ty, t的范围是[1e-3,1]，训练的时候t=1e-3对应的是x_0 * (1-eps)+ y*eps，但是推理的时候t的起点也是1e-3，这时候是没有y用于计算x_t的，是出于什么原因这样设计的呢？