is there a version for imbalance dataset?

[NickyTan8899]

is there a supervised contrastive loss for imbalance dataset just like the focal loss? This loss doesn't work when I use it for pretraining encoder with imbalance dataset.(the majority class works pretty well but the minority one is seriously worse)