feat: implement rae autoencoder. #13046
sayakpaul
merged
71 commits into
huggingface:main
from
Ando233:rae
feat: implement three RAE encoders(dinov2, siglip2, mae)
382aad0a
feat: finish first version of autoencoder_rae
f82cecc2
Merge branch 'main' into rae
a3926d77
Merge branch 'main' into rae
3ecf89d0
fix formatting
0850c8cd
make fix-copies
24acab0b
initial doc
25bc9e33
fix latent_mean / latent_var init types to accept config-friendly inputs
f06ea7a9
use mean and std convention
d7cb1247
cleanup
0d59b227
add rae to diffusers script
202b14f6
use imports
7cbbf271
use attention
e6d44993
remove unneeded class
6a9bde69
example traiing script
9522e68a
input and ground truth sizes have to be the same
906d79a4
fix argument
d3cbd5a6
move loss to training script
96520c4f
cleanup
fc529595
simplify mixins
a4fc9f64
fix training script
d06b5018
Merge branch 'main' into rae
d8b2983b
fix entrypoint for instantiating the AutoencoderRAE
c68b812c
added encoder_image_size config
61885f37
undo last change
28a02eb2
fixes from pretrained weights
b2978682
Merge branch 'main' into rae
7debd075
cleanups
b3ffd634
address reviews
dca59233
Merge branch 'rae' of https://github.com/Ando233/diffusers into rae
c71cb442
fix train script to use pretrained
5c857815
fix conversion script review
d965cabe
latebt normalization buffers are now always registered with no-op def…
663b5804
Update examples/research_projects/autoencoder_rae/README.md
6a787678
Update src/diffusers/models/autoencoders/autoencoder_rae.py
1b4a43f5
use image url
c717498f
Encoder is frozen
03e757ca
fix slow test
8861a808
remove config
9ada5768
use ModelTesterMixin and AutoencoderTesterMixin
34787e5b
make quality
cd86873e
Merge branch 'main' into rae
bff48d31
strip final layernorm when converting
f15873af
_strip_final_layernorm_affine for training script
e834e498
fix test
5817416a
add dispatch forward and update conversion script
3a668951
update training script
bc718898
Merge branch 'main' into rae
9c0f96b3
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
error out as soon as possible and add comments
196f8a36
Update src/diffusers/models/autoencoders/autoencoder_rae.py
3958fda3
use buffer
e805be98
inline
af75d8b9
Update src/diffusers/models/autoencoders/autoencoder_rae.py
df1af7d9
remove optional
876e9307
_noising takes a generator
b9a5266c
Update src/diffusers/models/autoencoders/autoencoder_rae.py
bc59324a
Merge branch 'main' into rae
14d918ee
dg845
commented
on 2026-03-04
dg845
commented
on 2026-03-04
fix api
33d98a85
rename
fa016b19
remove unittest
f4ec0f14
use randn_tensor
05d3edca
Merge branch 'huggingface:main' into rae
ed9bcfd7
Merge branch 'main' into rae
af0bed00
dg845
commented
on 2026-03-05
Merge branch 'main' into rae
441224ac
fix device map on multigpu
c0138b33
Merge branch 'rae' of https://github.com/Ando233/diffusers into rae
c640d01e
sayakpaul
commented
on 2026-03-05
sayakpaul
commented
on 2026-03-05
check if the key is missing in the original state dict and only then …
61e92205
Merge branch 'main' into rae
decac2ca
remove initialize_weights
87ee7fcf
sayakpaul
approved these changes
on 2026-03-05
Merge branch 'main' into rae
54cd3945
Merge branch 'main' into rae
e2fcc403
Login to write a write a comment.
Login via GitHub