Skip to content

Releases: lucidrains/equiformer-pytorch

0.3.10

08 Sep 08:53
Compare
Choose a tag to compare
address https://github.com/lucidrains/equiformer-pytorch/issues/12

0.3.9

27 Aug 18:54
Compare
Choose a tag to compare
oops

0.3.8

27 Aug 18:53
Compare
Choose a tag to compare
make sure to mask out padding tokens in mlp attention, at the attenti…

0.3.7

27 Aug 18:35
Compare
Choose a tag to compare
make sure padding tokens are never considered when ordering by relati…

0.3.6

23 Aug 14:53
Compare
Choose a tag to compare
some cleanup

0.3.5

23 Aug 14:16
Compare
Choose a tag to compare
actually modify relative distances for prioritizing adjacent nodes

0.3.4

22 Aug 16:10
Compare
Choose a tag to compare
small cleanup

0.3.3

22 Aug 15:32
Compare
Choose a tag to compare
port over edges and adjacency matrix logic

0.3.2

22 Jul 00:12
Compare
Choose a tag to compare
0.3.2

0.3.1

12 Jul 18:14
Compare
Choose a tag to compare
zero init the attention and feedforward branch outputs by default, bu…