GitLab

* fix: group offloading to support standalone computational layers in block-level offloading

* test: for models with standalone and deeply nested layers in block-level offloading

* feat: support for block-level offloading in group offloading config

* fix: group offload block modules to AutoencoderKL and AutoencoderKLWan

* fix: update group offloading tests to use AutoencoderKL and adjust input dimensions

* refactor: streamline block offloading logic

* Apply style fixes

* update tests

* update

* fix for failing tests

* clean up

* revert to use skip_keys

* clean up

---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com>

fix spatial compression ratio compute error for AutoEncoderKLWan
Co-authored-by: lirui.926 <lirui.926@bytedance.com>

* remove attn_processors property

* more

* up

* up more.

* up

* add AttentionMixin to AuraFlow.

* up

* up

* up

* up

* add


---------
Co-authored-by: yiyi@huggingface.co <yiyi@ip-26-0-161-123.ec2.internal>
Co-authored-by: yiyi@huggingface.co <yiyi@ip-26-0-160-103.ec2.internal>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

* add vae

* Initial commit for Flux 2 Transformer implementation

* add pipeline part

* small edits to the pipeline and conversion

* update conversion script

* fix

* up up

* finish pipeline

* Remove Flux IP Adapter logic for now

* Remove deprecated 3D id logic

* Remove ControlNet logic for now

* Add link to ViT-22B paper as reference for parallel transformer blocks such as the Flux 2 single stream block

* update pipeline

* Don't use biases for input projs and output AdaNorm

* up

* Remove bias for double stream block text QKV projections

* Add script to convert Flux 2 transformer to diffusers

* make style and make quality

* fix a few things.

* allow sft files to go.

* fix image processor

* fix batch

* style a bit

* Fix some bugs in Flux 2 transformer implementation

* Fix dummy input preparation and fix some test bugs

* fix dtype casting in timestep guidance module.

* resolve conflicts.,

* remove ip adapter stuff.

* Fix Flux 2 transformer consistency test

* Fix bug in Flux2TransformerBlock (double stream block)

* Get remaining Flux 2 transformer tests passing

* make style; make quality; make fix-copies

* remove stuff.

* fix type annotaton.

* remove unneeded stuff from tests

* tests

* up

* up

* add sf support

* Remove unused IP Adapter and ControlNet logic from transformer (#9)

* copied from

* Apply suggestions from code review
Co-authored-by: YiYi Xu <yixu310@gmail.com>
Co-authored-by: apolinário <joaopaulo.passos@gmail.com>

* up

* up

* up

* up

* up

* Refactor Flux2Attention into separate classes for double stream and single stream attention

* Add _supports_qkv_fusion to AttentionModuleMixin to allow subclasses to disable QKV fusion

* Have Flux2ParallelSelfAttention inherit from AttentionModuleMixin with _supports_qkv_fusion=False

* Log debug message when calling fuse_projections on a AttentionModuleMixin subclass that does not support QKV fusion

* Address review comments

* Update src/diffusers/pipelines/flux2/pipeline_flux2.py
Co-authored-by: YiYi Xu <yixu310@gmail.com>

* up

* Remove maybe_allow_in_graph decorators for Flux 2 transformer blocks (#12)

* up

* support ostris loras. (#13)

* up

* update schdule

* up

* up (#17)

* add training scripts (#16)

* add training scripts
Co-authored-by: Linoy Tsaban <linoytsaban@gmail.com>

* model cpu offload in validation.

* add flux.2 readme

* add img2img and tests

* cpu offload in log validation

* Apply suggestions from code review

* fix

* up

* fixes

* remove i2i training tests for now.

---------
Co-authored-by: Linoy Tsaban <linoytsaban@gmail.com>
Co-authored-by: linoytsaban <linoy@huggingface.co>

* up

---------
Co-authored-by: yiyixuxu <yixu310@gmail.com>
Co-authored-by: Daniel Gu <dgu8957@gmail.com>
Co-authored-by: yiyi@huggingface.co <yiyi@ip-10-53-87-203.ec2.internal>
Co-authored-by: dg845 <58458699+dg845@users.noreply.github.com>
Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com>
Co-authored-by: apolinário <joaopaulo.passos@gmail.com>
Co-authored-by: yiyi@huggingface.co <yiyi@ip-26-0-160-103.ec2.internal>
Co-authored-by: Linoy Tsaban <linoytsaban@gmail.com>
Co-authored-by: linoytsaban <linoy@huggingface.co>

Revert dim_mult back to list and fix type annotation

* Update pipeline_skyreels_v2_i2v.py

* Update README.md

* Update torch_utils.py

* Update torch_utils.py

* Update guider_utils.py

* Update pipeline_ltx.py

* Update pipeline_bria.py

* Apply suggestion from @qgallouedec

* Update autoencoder_kl_qwenimage.py

* Update pipeline_prx.py

* Update pipeline_wan_vace.py

* Update pipeline_skyreels_v2.py

* Update pipeline_skyreels_v2_diffusion_forcing.py

* Update pipeline_bria_fibo.py

* Update pipeline_skyreels_v2_diffusion_forcing_i2v.py

* Update pipeline_ltx_condition.py

* Update pipeline_ltx_image2video.py

* Update regional_prompting_stable_diffusion.py

* make style

* style

* style

Fix: update type hints for Tuple parameters across multiple files to support variable-length tuples (#12544)

* Fix: update type hints for Tuple parameters across multiple files to support variable-length tuples

* Apply style fixes

---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

* Changing the way we infer dtype to avoid force evaluation of lazy tensors

* changing way to infer dtype to ensure type consistency

* more robust infering of dtype

* removing the upscale dtype entirely

* fix crash in tiling mode is enabled
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* fmt
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

---------
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

* add hunyuanimage2.1


---------
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

add

* up

* correct wording.

* up

* up

* up

* change syntax

* make style

* Update autoencoder_kl_wan.py

When using the Wan2.2 VAE, the spatial compression ratio calculated here is incorrect. It should be 16 instead of 8. Pass it in directly via the config to ensure it’s correct here.

* Update autoencoder_kl_wan.py

* minor modification to support dc-ae-turbo

* minor

* fix(qwen-image):
- update vae license

* Apply style fixes

---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Aryan <aryan@huggingface.co>

* update

* update

* update

* add docs

* (feat): qwen-image integration

* fix(qwen-image):
- remove unused logics related to controlnet/ip-adapter

* fix(qwen-image):
- compatible with attention dispatcher
- cond cache support

* fix(qwen-image):
- cond cache registry
- attention backend argument
- fix copies

* fix(qwen-image):
- remove local test

* Update src/diffusers/models/transformers/transformer_qwenimage.py

---------
Co-authored-by: YiYi Xu <yixu310@gmail.com>

* Update autoencoder_kl_cosmos.py

* Apply style fixes

---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Aryan <aryan@huggingface.co>

up

* support wan 2.2 i2v

* add t2v + vae2.2

* add conversion script for vae 2.2

* add

* add 5b t2v

* conversion script

* refactor out reearrange

* remove a copied from in skyreels

* Apply suggestions from code review
Co-authored-by: bagheera <59658056+bghira@users.noreply.github.com>

* Update src/diffusers/models/transformers/transformer_wan.py

* fix fast tests

* style

---------
Co-authored-by: bagheera <59658056+bghira@users.noreply.github.com>

* update

* update

* update

update

* ⚡

️ Speed up method `AutoencoderKLWan.clear_cache` by 886%

**Key optimizations:**
- Compute the number of `WanCausalConv3d` modules in each model (`encoder`/`decoder`) **only once during initialization**, store in `self._cached_conv_counts`. This removes unnecessary repeated tree traversals at every `clear_cache` call, which was the main bottleneck (from profiling).
- The internal helper `_count_conv3d_fast` is optimized via a generator expression with `sum` for efficiency.

All comments from the original code are preserved, except for updated or removed local docstrings/comments relevant to changed lines.  
**Function signatures and outputs remain unchanged.**

* Apply style fixes

* Apply suggestions from code review
Co-authored-by: Aryan <contact.aryanvs@gmail.com>

* Apply style fixes

---------
Co-authored-by: codeflash-ai[bot] <148906541+codeflash-ai[bot]@users.noreply.github.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Aryan <aryan@huggingface.co>
Co-authored-by: Aryan <contact.aryanvs@gmail.com>
Co-authored-by: Aseem Saxena <aseem.bits@gmail.com>

* Fix typos in strings and comments
Signed-off-by: co63oc <co63oc@users.noreply.github.com>

* Update src/diffusers/hooks/hooks.py
Co-authored-by: Aryan <contact.aryanvs@gmail.com>

* Update src/diffusers/hooks/hooks.py
Co-authored-by: Aryan <contact.aryanvs@gmail.com>

* Update layerwise_casting.py

* Apply style fixes

* update

---------
Signed-off-by: co63oc <co63oc@users.noreply.github.com>
Co-authored-by: Aryan <contact.aryanvs@gmail.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

* Use HF Papers

* Apply style fixes

---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

* begin transformer conversion

* refactor

* refactor

* refactor

* refactor

* refactor

* refactor

* update

* add conversion script

* add pipeline

* make fix-copies

* remove einops

* update docs

* gradient checkpointing

* add transformer test

* update

* debug

* remove prints

* match sigmas

* add vae pt. 1

* finish CV* vae

* update

* update

* update

* update

* update

* update

* make fix-copies

* update

* make fix-copies

* fix

* update

* update

* make fix-copies

* update

* update tests

* handle device and dtype for safety checker; required in latest diffusers

* remove enable_gqa and use repeat_interleave instead

* enforce safety checker; use dummy checker in fast tests

* add review suggestion for ONNX export
Co-Authored-By: Asfiya Baig <asfiyab@nvidia.com>

* fix safety_checker issues when not passed explicitly

We could either do what's done in this commit, or update the Cosmos examples to explicitly pass the safety checker

* use cosmos guardrail package

* auto format docs

* update conversion script to support 14B models

* update name CosmosPipeline -> CosmosTextToWorldPipeline

* update docs

* fix docs

* fix group offload test failing for vae

---------
Co-authored-by: Asfiya Baig <asfiyab@nvidia.com>

* implement tiled encode/decode

* address review comments

* Add missing 'gradient_checkpointing = False' attr

* Add (limited) tests for Mochi autoencoder

* Apply style fixes

* pass 'conv_cache' as arg instead of kwarg

---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

* rewrite memory count without implicitly using dimensions by @ic-synth

* replace F.pad by built-in padding in Conv3D

* in-place sums to reduce memory allocations

* fixed trailing whitespace

* file reformatted

* in-place sums

* simpler in-place expressions

* removed in-place sum, may affect backward propagation logic

* removed in-place sum, may affect backward propagation logic

* removed in-place sum, may affect backward propagation logic

* reverted change

* update


---------
Co-authored-by: YiYi Xu <yixu310@gmail.com>
Co-authored-by: hlky <hlky@hlky.ac>