1.6.1

make sure rotary positional embeddings are same type as queries

1.6.0

add token shifting

1.5.2

apply rotary embeddings to values instead of using depthwise conv on …

1.5.1

fix rotary positional embedding

1.5.0

go all the way with rotary embeddings

1.4.3

allow for fine control over cancelling and registering kmeans update …

1.4.2

expose KmeansAttention layer for personal use in other repositories

1.4.1

fix examples

1.4.0

use embedding dimension, due to embedding factorization

1.3.0

allow one to look multiple blocks of window size into the past for lo…