Skip to content
GitLab
Explore
Sign in
Tags
Tags give the ability to mark specific points in history as being important
This project is mirrored from
https://github.com/lucidrains/vit-pytorch
. Pull mirroring updated
Sep 19, 2024
.
0.11.1
4b9a02d8
·
use depthwise conv for CvT projections
·
Mar 30, 2021
0.11.0
518924ea
·
add CvT
·
Mar 30, 2021
0.10.3
e712003d
·
add CrossViT
·
Mar 30, 2021
0.10.2
72c34f6b
·
add CrossViT
·
Mar 30, 2021
0.10.1
8135d70e
·
use hooks to retrieve attention maps for user without modifying ViT
·
Mar 29, 2021
0.9.3
171fd97a
·
remove masking, as it complicates with little benefit
·
Mar 23, 2021
0.9.2
2d413a39
·
remove masking, as it complicates with little benefit
·
Mar 23, 2021
0.9.1
b9008501
·
add deep vit
·
Mar 23, 2021
0.9.0
60346898
·
add deep vit
·
Mar 23, 2021
0.8.0
173e07e0
·
cleanup and release 0.8.0
·
Mar 08, 2021
0.7.6
1698b7be
·
make it so one can plug performer into t2tvit
·
Feb 25, 2021
0.7.5
be67c142
·
make it so one can plug performer into t2tvit
·
Feb 25, 2021
0.7.4
6760d554
·
no need to do projection to combine attention heads for T2Ts initial one-headed attention layers
·
Feb 24, 2021
0.7.3
dc7ee3e0
·
add DistillableT2TViT
·
Feb 21, 2021
0.7.2
c2e3d930
·
remove patch size from T2TViT
·
Feb 21, 2021
0.7.1
6af7bbcd
·
make sure distillation still works
·
Feb 21, 2021
0.7.0
bed61bec
·
add token to token ViT
·
Feb 19, 2021
0.6.8
3f2cbc6e
·
fix for ambiguity in broadcasting mask
·
Feb 17, 2021
0.6.7
85314cf0
·
patch for scaling factor, thanks to @urkax
·
Jan 21, 2021
0.6.5
e8ca6038
·
allow for DistillableVit to still run predictions
·
Jan 11, 2021
Prev
1
…
5
6
7
8
9
10
11
Next