Skip to content

Change KolorsPipeline LoRA Loader to StableDiffusion #11198

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Apr 3, 2025

Conversation

BasileLewan
Copy link
Contributor

The KolorsPipeline inherits from StableDiffusionXLLoraLoaderMixin which expects the pipeline to have two text encoders, whereas Kolors only uses one. This causes an error when trying to load LoRA weights on the pipeline.

Falling back to StableDiffusionLoraLoaderMixin fixes the issue.

(1st contribution here, sorry if I did something wrong)

Replace the SDXL LoRA Loader Mixin inheritance with the StableDiffusion one
@asomoza
Copy link
Member

asomoza commented Apr 2, 2025

oh nice, I forgot about that since at the time there weren't any kolors loras to test it. Can you please share the lora and an example of it?

@BasileLewan
Copy link
Contributor Author

I'm not sure to get what you mean, this is the case for any LoRA.
If you want a minimal reproduction, here it is:

import torch
from diffusers import KolorsPipeline
from diffusers.utils import convert_state_dict_to_diffusers
from peft import LoraConfig
from peft.utils import get_peft_model_state_dict

pipe = KolorsPipeline.from_pretrained("Kwai-Kolors/Kolors-diffusers", variant='fp16')
model = pipe.unet
cfg = LoraConfig(target_modules=["to_k", "to_q", "to_v"])

model.add_adapter(cfg)
lora_state_dict = convert_state_dict_to_diffusers(get_peft_model_state_dict(model))
pipe.save_lora_weights("lora_path", lora_state_dict)
pipe.load_lora_weights("lora_path/pytorch_lora_weights.safetensors")

# AttributeError: 'KolorsPipeline' object has no attribute 'text_encoder_2'

@asomoza
Copy link
Member

asomoza commented Apr 2, 2025

that example is nice but what I meant was a real lora trained that people use, the reason why this wasn't noticed by anyone until now except for you is that there wasn't any more people that used a Kolors Loras, even as unet loras, there's just a few and actually I never seen anyone using them.

Still doesn't matter, I was just curious, we still need to merge this in case that ever happens.

@HuggingFaceDocBuilderDev

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

@asomoza
Copy link
Member

asomoza commented Apr 3, 2025

I don't think there's a need to create a specific lora loader mixin for kolors since there are just a handful of them but this will enable people to use them if they want.

@asomoza asomoza closed this Apr 3, 2025
@asomoza asomoza reopened this Apr 3, 2025
@asomoza
Copy link
Member

asomoza commented Apr 3, 2025

thanks a lot! Failing tests are not related to this PR.

@asomoza asomoza merged commit 480510a into huggingface:main Apr 3, 2025
15 of 24 checks passed
@BasileLewan
Copy link
Contributor Author

No problem, thanks for the quick review!

jonluca added a commit to weights-ai/diffusers that referenced this pull request Apr 3, 2025
* Raise warning and round down if Wan num_frames is not 4k + 1 (huggingface#11167)

* update

* raise warning and round to nearest multiple of scale factor

* [Docs] Fix environment variables in `installation.md` (huggingface#11179)

* Add `latents_mean` and `latents_std` to `SDXLLongPromptWeightingPipeline` (huggingface#11034)

* Bug fix in LTXImageToVideoPipeline.prepare_latents() when latents is already set (huggingface#10918)

* Bug fix in ltx

* Assume packed latents.

---------

Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com>
Co-authored-by: YiYi Xu <yixu310@gmail.com>

* [tests] no hard-coded cuda  (huggingface#11186)

no cuda only

* [WIP] Add Wan Video2Video (huggingface#11053)

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* map BACKEND_RESET_MAX_MEMORY_ALLOCATED to reset_peak_memory_stats on XPU (huggingface#11191)

Signed-off-by: YAO Matrix <matrix.yao@intel.com>

* fix autocast (huggingface#11190)

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix: for checking mandatory and optional pipeline components (huggingface#11189)

fix: optional componentes verification on load

* remove unnecessary call to `F.pad` (huggingface#10620)

* 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

* allow models to run with a user-provided dtype map instead of a single dtype (huggingface#10301)

* allow models to run with a user-provided dtype map instead of a single dtype

* make style

* Add warning, change `_` to `default`

* make style

* add test

* handle shared tensors

* remove warning

---------

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

* [tests] HunyuanDiTControlNetPipeline inference precision issue on XPU (huggingface#11197)

* add xpu part

* fix more cases

* remove some cases

* no canny

* format fix

* Revert `save_model` in ModelMixin save_pretrained and use safe_serialization=False in test (huggingface#11196)

* [docs] `torch_dtype` map (huggingface#11194)

* Fix enable_sequential_cpu_offload in CogView4Pipeline (huggingface#11195)

* Fix enable_sequential_cpu_offload in CogView4Pipeline

* make fix-copies

* SchedulerMixin from_pretrained and ConfigMixin Self type annotation (huggingface#11192)

* Update import_utils.py (huggingface#10329)

added onnxruntime-vitisai for custom build onnxruntime pkg

* Add CacheMixin to Wan and LTX Transformers (huggingface#11187)

* update

* update

* update

* feat: [Community Pipeline] - FaithDiff Stable Diffusion XL Pipeline (huggingface#11188)

* feat: [Community Pipeline] - FaithDiff Stable Diffusion XL Pipeline for Image SR.

* added pipeline

* [Model Card] standardize advanced diffusion training sdxl lora (huggingface#7615)

* model card gen code

* push modelcard creation

* remove optional from params

* add import

* add use_dora check

* correct lora var use in tags

* make style && make quality

---------

Co-authored-by: Aryan <aryan@huggingface.co>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

* Change KolorsPipeline LoRA Loader to StableDiffusion (huggingface#11198)

Change LoRA Loader to StableDiffusion

Replace the SDXL LoRA Loader Mixin inheritance with the StableDiffusion one

* Update Style Bot workflow (huggingface#11202)

update style bot workflow

---------

Signed-off-by: YAO Matrix <matrix.yao@intel.com>
Signed-off-by: jiqing-feng <jiqing.feng@intel.com>
Co-authored-by: Aryan <aryan@huggingface.co>
Co-authored-by: Mark <remarkablemark@users.noreply.github.com>
Co-authored-by: hlky <hlky@hlky.ac>
Co-authored-by: kakukakujirori <63725741+kakukakujirori@users.noreply.github.com>
Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com>
Co-authored-by: YiYi Xu <yixu310@gmail.com>
Co-authored-by: Fanli Lin <fanli.lin@intel.com>
Co-authored-by: Yao Matrix <matrix.yao@intel.com>
Co-authored-by: jiqing-feng <jiqing.feng@intel.com>
Co-authored-by: Eliseu Silva <elismasilva@gmail.com>
Co-authored-by: Bruno Magalhaes <bruno.magalhaes@synthesia.io>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
Co-authored-by: lakshay sharma <31830611+Lakshaysharma048@users.noreply.github.com>
Co-authored-by: Abhipsha Das <ad6489@nyu.edu>
Co-authored-by: Basile Lewandowski <basile.lewan@gmail.com>
Co-authored-by: célina <hanouticelina@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants