search

jonbarron / robust_loss_pytorch

A pytorch port of google-research/google-research/robust_loss/
Apache License 2.0
656 stars 88 forks source link

AdaptiveLossFunction question #31

Closed 110dazhi closed 2 years ago

110dazhi commented 2 years ago

when I use the same target and the same predict, like "target = a" and "predict=a", I find the AdaptiveLossFunction is not equal 0. Is it normal? adative

jonbarron commented 2 years ago

Yep that makes sense, the adaptive loss is a negative log likelihood, so it won't necessarily bottom out at 0.