Please leave the trained model for me, thanks a lot!

You implement loss = self.alpha * (1-y) * distance**2 + \ self.beta * y * (torch.max(torch.zeros_like(distance), self.margin - distance)**2) as your contrastive loss, however, in your dataset split and preprocessing script, you label (genuine, genuine) as 1 and (genuine,forged) as 0, which means when y=0, your loss = alpha * distance between pairs and it will be minimized, but hopefully they should be as far as possible.

loss = self.alpha * y * distance**2 + \ self.beta * (1-y) * (torch.max(torch.zeros_like(distance), self.margin - distance)**2) would be a correct implementation.

The last layer is 256, but the SigNet paper specifies 128.
(Note: the network trains much more quickly with 128.)

In train.py after print the result for any log_interval step why set number_samples and running_loss to zero ?

But in the eval step do not set them to 0.

我看原论文中模型里所有的最大池化层大小都是33，在这个实现里都是22的，请问有没有试验过两种哪个性能好呢？