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/video-diffusion-pytorch
. Pull mirroring updated
Sep 19, 2024
.
0.1.0
c1fcc03a
·
save samples as one big grid of gifs
·
Apr 21, 2022
0.0.20
60c796e9
·
oops
·
Apr 20, 2022
0.0.19
6a61ecbb
·
convert images to tensor from range of -1 to 1, unnormalize before saving gif at sample time
·
Apr 20, 2022
0.0.18
3925bac5
·
add an assert
·
Apr 20, 2022
0.0.17
2ff61f55
·
bump
·
Apr 20, 2022
0.0.16
dcf67c1d
·
make sure one can train on gifs of channel 1, and be able to specify number of frames other than 16
·
Apr 20, 2022
0.0.14
2599ae21
·
complete simple trainer class for folder of gifs (for starters) for now,...
·
Apr 20, 2022
0.0.12
c795da23
·
fix cosine beta schedule
·
Apr 19, 2022
0.0.11
90807a1d
·
fix unconditional unet3d
·
Apr 18, 2022
0.0.10
a93c8c83
·
fix bug made in translation of numpy to torch
·
Apr 15, 2022
0.0.9
49c0e0da
·
offer means for attending to the present
·
Apr 14, 2022
0.0.8
f8b79163
·
complete relative positional encoding for attention across time through attention bias
·
Apr 13, 2022
0.0.7
72d113c3
·
make it even easier
·
Apr 13, 2022
0.0.6
0d8f175a
·
add Attention across Time after all (spatial) linear attention
·
Apr 13, 2022
0.0.4
174a8967
·
actually use classifier free guidance during sampling, using the cond_scale > 1
·
Apr 13, 2022
0.0.2
12fac4c8
·
attend to space and time
·
Apr 13, 2022
0.0.1
0c198b11
·
modify unet to be all conv3ds, complete all spatial attention, but temporal...
·
Apr 13, 2022
Prev
1
2
3
Next