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/make-a-video-pytorch
. Pull mirroring updated
Sep 19, 2024
.
0.4.0
2f575461
·
remove groupnorms
https://arxiv.org/abs/2312.02696
·
May 03, 2024
0.3.1
16561b3e
·
ability to do causal attention for time axis
·
Jun 19, 2023
0.3.0
f2d214f5
·
add flash attention
·
Jun 19, 2023
0.2.1
57ed18b3
·
feedforward should be agnostic to images or video
·
Jun 19, 2023
0.2.0
4430a87d
·
fix layer norm in feedforward
·
Mar 22, 2023
0.1.2
3280b3d4
·
token shift within feedforward best accompanied by a layernorm, from normformer paper
·
Mar 19, 2023
0.1.1
bde9f228
·
quick fix
·
Mar 19, 2023
0.1.0
7e0838e3
·
follow spatiotemporal attention with a feedforward, and add the highly...
·
Mar 19, 2023
0.0.10
b6e0a17c
·
make continuous positional bias calculations more efficient, while...
·
Mar 18, 2023
0.0.9
896a4d03
·
allow for customizing number of resnet blocks per stage in space time unet
·
Dec 13, 2022
0.0.8
5a93a28d
·
exercise attention expertise
·
Dec 10, 2022
0.0.7
e634856e
·
2 skip connections per stage, and also make sure properly init upsample pixelshuffle
·
Dec 10, 2022
0.0.6
58f4172a
·
some asserts
·
Dec 10, 2022
0.0.5
8de344e0
·
now accepts timestep conditioning, for ddpm
·
Dec 10, 2022
0.0.4
67565c1d
·
skip connections
·
Dec 10, 2022
0.0.3
a4c37ee7
·
a full space time unet where time can be ignored
·
Dec 10, 2022
0.0.2
0af63d7d
·
it all clicked in my head
·
Dec 09, 2022
0.0.1
225f3047
·
add the main contributions of the paper
·
Oct 17, 2022