0.3.0

fix so that encoding of retrieved tokens happen just before the first…

0.2.10

0.2.10

0.2.9

Update setup.py

0.2.8

allow for use of gated RMSNorm, proposed by RMSNorm author in the ReL…

0.2.7

rename hyperparameter

0.2.6

complete deepnet init

0.2.5

default to layernorm if post norm configuration is turned on, but all…

0.2.4

separate projection matrices for key / values, in preparation for sep…

0.2.3

default residual scaling to equations proposed in deepnet paper, if p…

0.2.2

allow for post norm configuration for deepnet + retro