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/magvit2-pytorch
. Pull mirroring updated
Sep 19, 2024
.
0.0.27
72a86e82
·
add linear space attention layer, using a specific one from meta ai
·
Oct 24, 2023
0.0.26
3cc44f7d
·
make tokenizer conditionable
·
Oct 23, 2023
0.0.25
eaba56f8
·
use rmsnorm for transformer blocks, and also add normformer, as @borisdayma...
·
Oct 22, 2023
0.0.23
2e1deaaa
·
setup the residual unit, but with modulatable conv3d
·
Oct 22, 2023
0.0.22
930367b3
·
bring in token shifting along the time axis
·
Oct 22, 2023
0.0.21
3a7200be
·
follow attention with feedforward
·
Oct 19, 2023
0.0.19
96b52aa7
·
make conv3d mod causality optional
·
Oct 19, 2023
0.0.18
f15355ef
·
attend to space and time (causally)
·
Oct 19, 2023
0.0.17
433f5843
·
quick fix for training on images
·
Oct 19, 2023
0.0.16
49dd0e93
·
residual output is always input dim
·
Oct 19, 2023
0.0.15
645d6838
·
forward all loss breakdown from LFQ
·
Oct 19, 2023
0.0.14
46173c0d
·
be able to weigh the lookup free quantizer auxiliary loss
·
Oct 19, 2023
0.0.12
408dfcdb
·
weigh the generator loss
·
Oct 18, 2023
0.0.11
e1540313
·
add antialiasing in discriminator, project management
·
Oct 18, 2023
0.0.10
0a9f7378
·
complete the GAN component
·
Oct 18, 2023
0.0.9
03ac452a
·
start adding the gan components, starting with discriminator loss + gradient penalty for stability
·
Oct 18, 2023
0.0.8
4fe048e5
·
allow for antialiasing strided downsamples in video tokenizer for starters
·
Oct 17, 2023
0.0.7
d2f2406a
·
prepare blur for blurpool, bring in the discriminator being used successfully at phenaki-pytorch
·
Oct 17, 2023
0.0.6
1b56b7b9
·
add perceptual loss and some project management
·
Oct 16, 2023
0.0.5
5a5ad490
·
update to latest
·
Oct 16, 2023
Prev
1
2
3
4
5
6
7
Next