GitLab

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

fix
Co-authored-by: yiyixuxu <yixu310@gmail,com>

* Add attention masking to attn processors

* Update tensor conversion

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

* make text encoder component truly optional.

* more fixes

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

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

Fixed deprecation warning for torch.utils._pytree._register_pytree_node in PyTorch 2.2
Co-authored-by: Yinghua <yzho0423@uni.sydney.edu.au>

* fix

* update docstring

---------
Co-authored-by: yiyixuxu <yixu310@gmail,com>

* Fixed typos in __init__ and in forward of Unet3DConditionModel

* Resolving conflicts

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

copy docstring for `strength` from  stablediffusion img2img pipeline to controlnet img2img pipelines

* Fix words

* Fix

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

* feat: allow low_cpu_mem_usage in ip adapter loading

* reduce the number of device placements.

* documentation.

* throw low_cpu_mem_usage warning only once from the main entry point.

* use cases

* first draft

* fix image links

* lcm-lora

* feedback

* review

* feedback

* feedback

* use load_model_into_meta in single file utils

* propagate to autoencoder and controlnet.

* correct class name access behaviour.

* remove torch_dtype from load_model_into_meta; seems unncessary

* remove incorrect kwarg

* style to avoid extra unnecessary line breaks

* update

* update

* update

[DPMSolverSinglestepScheduler] correct `get_order_list` for `solver_order=2`and `lower_order_final=True`  (#6953)

* add

* change default

---------
Co-authored-by: yiyixuxu <yixu310@gmail,com>

* update

fix

* fix: bias loading bug

* fixes for SDXL

* apply changes to the conversion script to match single_file_utils.py

* do transpose to match the single file loading logic.

* update

* update

* update

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

* fix

* add tests

---------
Co-authored-by: yiyixuxu <yixu310@gmail,com>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

add
Co-authored-by: yiyixuxu <yixu310@gmail,com>

* post release

* style

* Empty-Commit

---------
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* deprecate certain lora methods from the old backend.

* uncomment necessary things.

* safe remove old lora backend 👋

* refactor decode latents in video pipelines

* make fix-copies

* Add vae tiling in img2img and inpaint

* Add vae tiling not slicing

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

change to 2024

This fixes #6899.

* attention_head_dim

* debug

* print more info

* correct num_attention_heads behaviour

* down_block_num_attention_heads -> num_attention_heads.

* correct the image link in doc.

* add: deprecation for num_attention_head

* fix: test argument to use attention_head_dim

* more fixes.

* quality

* address comments.

* remove depcrecation.

* add: support for passing ip adapter image embeddings

* debugging

* make feature_extractor unloading conditioned on safety_checker

* better condition

* type annotation

* index to look into value slices

* more debugging

* debugging

* serialize embeddings dict

* better conditioning

* remove unnecessary prints.

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

* make fix-copies and styling.

* styling and further copy fixing.

* fix: check_inputs call in controlnet sdxl img2img pipeline

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

* remove torch_dtype from to()

* remove torch_dtype from usage scripts.

* remove old lora backend

* Revert "remove old lora backend"

This reverts commit adcddf6ba421f847e7da2a0ce57b9456cae43356.

[ip-adapter] make sure length of `scale` is same as number of ip-adapters when using  `set_ip_adapter_scale` (#6884)

add
Co-authored-by: yiyixuxu <yixu310@gmail,com>

* feat: standarize model card creation for dreambooth training.

* correct 'inference

* remove comments.

* take component out of kwargs

* style

* add: card template to have a leaner description.

* widget support.

* propagate changes to train_dreambooth_lora

* propagate changes to custom diffusion

* make widget properly type-annotated

This reverts commit 15f6b224.

* remove _to_tensor

* remove _to_tensor definition

* remove _collapse_frames_into_batch

* remove lora for not bloating the code.

* remove sample_size.

* simplify code a bit more

* ensure timesteps are always in tensor.

* Fix `AutoencoderTiny` with `use_slicing`

When using slicing with AutoencoderTiny, the encoder mistakenly encodes the entire batch for every image in the batch.

* Fixed formatting issue