0.2.8

handle if even amount of batch sizes across devices

0.2.7

export the sigmoid contrastive learning module

0.2.6

further cleanup the distributed code

0.2.5

quick fix

0.2.4

complete distributed logic for sigmoid contrastive loss

0.2.3

make sure regular contrastive loss supports distributed data parallel…

0.2.2

move beartype onto methods

0.2.1

missing import

0.2.0

add sigmoid contrastive loss

0.1.2

contrastive loss needs both directions