Why the loss is different from BYOL authors'

[Jing-XING]

I found the loss is different from the loss said in BYOL paper which should be a L2 loss and I did't find explanation... The loss in this repo is a cosine loss, and I just want to know why. BTW, thanks for this great repo!

[lucidrains]

if you read section J.3 in the paper, the code is identical

[Jing-XING]

if you read section J.3 in the paper, the code is identical

Thanks for your reply! I see, after normalization the L2 loss is the same as cosine similarity