0.2.11

switch to local attention

0.2.10

make sure linear attentino is paired with a feedforward, place linear…

0.2.9

fix a bug with is_neighbor_mask (for auto determining face edges from…

0.2.8

handle linear attention and channel first inputs in decoder

0.2.7

oops, thanks Marcus

0.2.6

remove need for padding computation for graph conv using masked_selec…

0.2.5

remove need for padding computation for graph conv using masked_selec…

0.2.4

able to customize attn kwargs for linear attention autoencoder, defau…

0.2.3

able to customize attn kwargs for linear attention autoencoder, defau…

0.2.2

allow for temperature of 0, for testing kv caching, which is more com…