Pytorch implementation of LNS 

LNS provides a very effiecient way for adaptative-hard negative sampling. I'm very interested in its use in contrastive learning applications. Do you think it is possible to implement LNS in pytorch or tensorflow? I did not figure it out since hashing seems not work in batch tensor computation, especially when each buckets have different number of samples. Many thanks!