1.42.4

move value residual so that kv caching works properly

1.42.3

fix value residual

1.42.2

maintain same behavior as an mlp when no batch dim

1.42.0

attention is all you need for neural radiance fields

1.41.5

address https://github.com/lucidrains/x-transformers/issues/280

1.41.3

add ability to scale log of forget decay in forgetting transformers, …

1.41.2

add an improvised per row data dependent alibi, using attention-esque…

1.41.1

follow some suggestions from @faresobeid

1.41.0

add data dependent alibi from the forgetting transformers paper from …

1.40.11

multiple cls tokens support from @pradeep-pyro