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
.
0.1.23
e3bab3d4
·
0.1.23
·
Nov 16, 2023
0.1.22
8de12e3c
·
autocast context for accelerator
·
Nov 16, 2023
0.1.21
b3a7c146
·
remove a gotcha for ddp training
·
Nov 16, 2023
0.1.20
e1a0ecfc
·
allow for encoding the first frame separate from the rest, like in cvivit
·
Nov 16, 2023
0.1.19
8182124e
·
address
https://github.com/lucidrains/magvit2-pytorch/issues/9
·
Nov 16, 2023
0.1.18
a07a9404
·
0.1.18
·
Nov 14, 2023
0.1.17
33d949c0
·
0.1.17
·
Nov 13, 2023
0.1.16
bcbe36a4
·
0.1.16
·
Nov 06, 2023
0.1.15
ea28dde1
·
add a final norm after all the encoder layers, just before quantizing
·
Nov 03, 2023
0.1.14
681d0b35
·
allow for multiple number of codebooks, so FSQ can be fairly evaluated against LFQ
·
Nov 03, 2023
0.1.12
e0719491
·
will just use x-transformers for the downstream (non)autoregressive model
·
Nov 03, 2023
0.1.11
3acd00b2
·
fix mishandling of ema model device
·
Nov 03, 2023
0.1.10
266c4ba1
·
force ema to be on same device as online model for now before update
·
Nov 02, 2023
0.1.9
234680d6
·
fix a device issue with ema model
·
Nov 02, 2023
0.1.8
ebee0725
·
log both online and EMA validation recon losses
·
Nov 02, 2023
0.1.7
ad1a799c
·
log the validation recon loss
·
Nov 02, 2023
0.1.6
29ae48c0
·
be able to specify run name when using trackers contextmanager, just something...
·
Nov 02, 2023
0.1.5
6336c829
·
init residual branch to zero
·
Nov 02, 2023
0.1.4
32c802b9
·
no blurring in main tokenizer network, only in discriminator
·
Nov 02, 2023
0.1.2
d4fb30d3
·
setup weights and biases tracking
·
Nov 02, 2023
Prev
1
2
3
4
5
6
7
Next