Loss function always return 0

[hxue3]

I am running into an issue that the loss function always returns 0 while using triplet loss function.

And if I use cosine-softmax, it won't return 0 but it will give me another error LossTensor is inf or nan : Tensor had NaN values

Any idea why