Closed kkokilep closed 1 year ago
Why does all the implementations seem to come with a class loss in addition to the loss associated with the method? I don't understand the reasoning for this. Any help with this issue is appreciated!
It is just for evaluation purposes, the linear layer receives features that are detached. See previous discussion: https://github.com/vturrisi/solo-learn/issues/230
Why does all the implementations seem to come with a class loss in addition to the loss associated with the method? I don't understand the reasoning for this. Any help with this issue is appreciated!