0.3.7

start moving eng stuff to another package

0.3.6

return loss breakdown and expose reconstruction loss

0.3.5

able to make model dimensions of fine vertex transformer to be smalle…

0.3.4

address a bug with auto appending eos, thanks to @qixuema

0.3.3

address https://github.com/lucidrains/meshgpt-pytorch/issues/14

0.3.2

make gateloop cache work

0.3.1

default kv cache to true on generate

0.3.0

finish fine transformer kv caching, 7x speedup now

0.2.25

first take care of caching on the coarse transformer, 3x speedup

0.2.24

remove text related keys returned from dataset, for sharing same data…