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
.
0.0.49
1b45835e
·
remove a gotcha where residual dropout is in effect when fetching codes....
·
Oct 28, 2023
0.0.48
9e590c11
·
tweak and readme
·
Oct 28, 2023
0.0.47
f30cbfb2
·
complete saving video and reconstructed video to results folder during validation
·
Oct 28, 2023
0.0.46
3952eebc
·
should be able to examine all the auxiliary losses under eval mode
·
Oct 28, 2023
0.0.45
d3ef4b6f
·
add consecutive residual layer
·
Oct 27, 2023
0.0.44
103bbc32
·
ema model should only be updated on main
·
Oct 27, 2023
0.0.43
f22c3d20
·
saving checkpoint
·
Oct 27, 2023
0.0.42
6146ee7d
·
saving and loading trainer
·
Oct 27, 2023
0.0.41
f4251c03
·
setup basic recon validation
·
Oct 27, 2023
0.0.40
f6bcdd71
·
better initial scale for xcit linear attention
·
Oct 26, 2023
0.0.39
45ef9b44
·
latest accelerate
·
Oct 26, 2023
0.0.38
c42c342e
·
fully wire up accelerate
·
Oct 26, 2023
0.0.37
9db0ec98
·
preliminary training loop
·
Oct 26, 2023
0.0.36
c3d41597
·
just copy over a optimizer factory fn
·
Oct 26, 2023
0.0.35
0fd20380
·
apply transform to each frame of video
·
Oct 26, 2023
0.0.34
edbddc2c
·
bring in dataset and dataloader
·
Oct 26, 2023
0.0.33
6dfc98cb
·
saves the video tokenizer config in the .pt file
·
Oct 25, 2023
0.0.31
b3d4644c
·
save version
·
Oct 25, 2023
0.0.30
cce7d8f8
·
add linear attention blocks to discriminator
·
Oct 24, 2023
0.0.28
14047c96
·
assert that decoding from code indices works correctly and is the same as...
·
Oct 24, 2023
Prev
1
2
3
4
5
6
7
Next