0.0.12

start using a specific linear attention for channel modulation, which…

0.0.11

redo the hrrr norm strategy to be completely flexible

0.0.10

redo the hrrr norm strategy to be completely flexible

0.0.9

first pass at inputs, redo hrrr

0.0.8

make sure targets and losses are to spec

0.0.7

offer another way to normalize hrrr targets and predictions

0.0.6

running mean and variance should only be updated during training

0.0.5

normalization of hrrr pred and targets should use same statistics

0.0.4

normalization of hrrr pred and targets should use same statistics

0.0.3

use a sync batchnorm as way of normalizing all channels of the HRRR, …