0.1.0

add sub-groupnorm in multihead ema, for use in audio modeling

0.0.15

fix laplace activation function thanks to @boweny-cerebras

0.0.14

fix residual within mega layer, thanks to @VHellendoorn

0.0.12

prenorm requires a final layernorm

0.0.11

fix residual for prenorm

0.0.10

expose multi-headed learned EMA, for use outside of repo

0.0.9

offer prenorm architecture

0.0.8

handle bidirectional better

0.0.7

Full Changelog: 0.0.6...0.0.7

0.0.6

improvise on bidirectional for multi-head learned ema