NCL implementation

[seungyeon-seo]

In the paper, NCL loss comprises loss for augmented data and loss for pseudo-positives. I understand the implementation in NCL.py file. In the forward function, you regard the origin data and its augmentation as positives, and all the data in queue M as negatives. The paper describes that the loss of augmented data and the loss of pseudo-positives have to be combined by alpha. However, you combine the first term of NCL (loss of augmented data) and SCL (loss_ncl_l) using alpha. Where is the second term of NCL (for pseudo-positive)?