Has anyone tried adding self-supervised constrative loss to SCL?

HobbitLong / SupContrast

PyTorch implementation of "Supervised Contrastive Learning" (and SimCLR incidentally)

BSD 2-Clause "Simplified" License

3.09k stars 530 forks source link

Open Chen-Song opened 4 years ago

Chen-Song commented 4 years ago

That is to say, supervised contrastive loss and self-supervised constrative loss, the former can be regarded as global and the latter is local.

adv010 commented 1 year ago

Hi @Chen-Song , did you try this out?