let's go Flux2 🚀 #12711
sayakpaul
merged
76 commits into
main
from
full-flux2
add vae
fce87b36
Initial commit for Flux 2 Transformer implementation
a1f2ba1a
add pipeline part
e470643e
small edits to the pipeline and conversion
7456a498
Merge pull request #3 from huggingface/flux2-small-edits
f3405998
update conversion script
b10e6296
fix
73823584
Merge branch 'flux2' of github.com:huggingface/diffusers-new-model-ad…
78ea69d6
up up
524b1238
finish pipeline
8a48adcd
Remove Flux IP Adapter logic for now
429d2cf5
Remove deprecated 3D id logic
2d7bad73
Remove ControlNet logic for now
c67f5820
Add link to ViT-22B paper as reference for parallel transformer block…
7acd7dad
update pipeline
66040866
Don't use biases for input projs and output AdaNorm
a200780f
up
f7148a02
Remove bias for double stream block text QKV projections
54c60803
Add script to convert Flux 2 transformer to diffusers
dba4c1fe
make style and make quality
cceffc4b
fix a few things.
e93a7461
allow sft files to go.
cfdd0057
Merge pull request #4 from huggingface/flux2-model-cpu
edacb522
fix image processor
14839807
fix batch
62231ef9
style a bit
68db1782
Merge pull request #5 from huggingface/small-flux2-transformer-fix
89e42d94
Fix some bugs in Flux 2 transformer implementation
f6c82a31
Fix dummy input preparation and fix some test bugs
4082c430
fix dtype casting in timestep guidance module.
2b5b2e33
Merge pull request #6 from huggingface/fix-dtype
247770fd
resolve conflicts.,
3d022b8b
remove ip adapter stuff.
ffb00619
Fix Flux 2 transformer consistency test
6820d6ca
Fix bug in Flux2TransformerBlock (double stream block)
f6059b75
merge
5a108f0c
Get remaining Flux 2 transformer tests passing
05fcddbf
make style; make quality; make fix-copies
ac4f61a1
remove stuff.
20f9b830
Merge branch 'flux2-transformer' into full-flux2
e8d71ab4
fix type annotaton.
546e60da
remove unneeded stuff from tests
eeb52c28
tests
771e17bc
up
28da6799
up
14f98663
add sf support
6d169759
Remove unused IP Adapter and ControlNet logic from transformer (#9)
be6604bb
copied from
ec0a2add
Apply suggestions from code review
980bcc86
up
e1d46ce4
up
de993441
up
739c31b8
up
ded81a9c
up
c0086ea1
Refactor Flux2Attention into separate classes for double stream and s…
4f1f67a8
Add _supports_qkv_fusion to AttentionModuleMixin to allow subclasses …
b8e37600
Have Flux2ParallelSelfAttention inherit from AttentionModuleMixin wit…
24159acc
Log debug message when calling fuse_projections on a AttentionModuleM…
cb21400c
Address review comments
ba52a591
Merge pull request #11 from huggingface/fix-fused-projections
cc74755d
Update src/diffusers/pipelines/flux2/pipeline_flux2.py
697a43c9
up
aba48dde
Remove maybe_allow_in_graph decorators for Flux 2 transformer blocks …
963ed57b
up
997dfc24
Merge branch 'main' into full-flux2
842a8496
support ostris loras. (#13)
8d8bb3da
up
454cef8d
Merge branch 'main' into full-flux2
6bf955aa
update schdule
fc1bd891
up
29b02b80
Merge pull request #15 from huggingface/final-update
24ff10e3
Merge pull request #14 from huggingface/transformer-single-file-support
4ef0285b
up (#17)
823f4c3f
add training scripts (#16)
5419877c
up
0a5a4773
Merge branch 'main' into full-flux2
cb9f1246
yiyixuxu
approved these changes
on 2025-11-25
DN6
approved these changes
on 2025-11-25
Login to write a write a comment.
Login via GitHub