27 Pull requests merged by 13 people

• [LoRA] small change to support Hunyuan LoRA Loading for FramePack

  #11546 merged May 13, 2025

• [train_dreambooth.py] Fix the LR Schedulers when num_train_epochs is passed in a distributed training env

  #11239 merged May 13, 2025

• fix: remove torch_dtype="auto" option from docstrings

  #11513 merged May 13, 2025

• Fix typo in train_diffusion_orpo_sdxl_lora_wds.py

  #11541 merged May 13, 2025

• Add VisualCloze

  #11377 merged May 12, 2025

• [test_models_transformer_ltx.py] help us test torch.compile() for impactful models

  #11512 merged May 12, 2025

• enable several pipeline integration tests on XPU

  #11526 merged May 12, 2025

• Hunyuan Video Framepack F1

  #11534 merged May 12, 2025

• [tests] add tests for framepack transformer model.

  #11520 merged May 11, 2025

• Change Framepack transformer layer initialization order

  #11535 merged May 9, 2025

• enable print_env on xpu

  #11507 merged May 9, 2025

• enable dit integration cases on xpu

  #11523 merged May 9, 2025

• [LTXPipeline] Update latents dtype to match VAE dtype

  #11533 merged May 9, 2025

• enable 7 cases on XPU

  #11503 merged May 9, 2025

• [LoRA] support non-diffusers hidream loras

  #11532 merged May 9, 2025

• [Tests] Enable more general testing for torch.compile() with LoRA hotswapping

  #11322 merged May 9, 2025

• feat: pipeline-level quantization config

  #11130 merged May 9, 2025

• [tests] fix audioldm2 for transformers main.

  #11522 merged May 8, 2025

• Conditionally import torchvision in Cosmos transformer

  #11524 merged May 8, 2025

• Add cross attention type for Sana-Sprint training in diffusers.

  #11514 merged May 8, 2025

• [LoRA] make lora alpha and dropout configurable

  #11467 merged May 8, 2025

• clean up the __Init__ for stable_diffusion

  #11500 merged May 7, 2025

• Cosmos

  #10660 merged May 7, 2025

• [docs] minor updates to bitsandbytes docs.

  #11509 merged May 6, 2025

• [lora_conversion] Enhance key handling for OneTrainer components in LORA conversion utility (#11441)

  #11487 merged May 6, 2025

• enable lora cases on XPU

  #11506 merged May 6, 2025

• Hunyuan Video Framepack

  #11428 merged May 6, 2025

14 Pull requests opened by 12 people

• [tests] Changes to the `torch.compile()` CI and tests

  #11508 opened May 6, 2025

• Add FluxPAGPipeline with support for PAG

  #11510 opened May 6, 2025

• [tests] add test for group_offloading with training.

  #11515 opened May 7, 2025

• LTX Video 0.9.7

  #11516 opened May 7, 2025

• Feat/multispectral dataloader 5ch

  #11517 opened May 7, 2025

• Add SkyReels-V2 pipelines

  #11518 opened May 7, 2025

• RegionalPrompting: Inherit from Stable Diffusion

  #11525 opened May 8, 2025

• enable pipeline test cases on xpu

  #11527 opened May 9, 2025

• Doc update

  #11531 opened May 9, 2025

• [tests] Add torch.compile test for UNet2DConditionModel

  #11537 opened May 10, 2025

• Fix deprecation warnings in test_ltx_image2video.py

  #11538 opened May 10, 2025

• [WIP] Modular Diffusers support custom code/pipeline blocks

  #11539 opened May 12, 2025

• [Chore] add rule to ignore checking black line whitespace.

  #11544 opened May 13, 2025

• Update pipeline_flux_img2img.py to add missing vae_slicing and vae_tiling calls.

  #11545 opened May 13, 2025

14 Issues closed by 6 people

• `.from_pretrained` `torch_dtype="auto"` argument not working a expected

  #11432 closed May 13, 2025

• wan2.1 transformer gguf load error

  #11088 closed May 13, 2025

• CompVis/stable-diffusion-v1-4 is missing fp16 files

  #11295 closed May 12, 2025

• Prompt adherence for FluxPipeline is broken

  #11536 closed May 10, 2025

• [Tracker] fix training resuming problem when using FP16 in the examples

  #6552 closed May 10, 2025

• [Feature request] LTX-Video v0.9.6 15x faster inference than non-distilled model.

  #11359 closed May 9, 2025

• HiDream LoRA support

  #11383 closed May 9, 2025

• Apply applicable `quantization_config` to model components when loading a model

  #10327 closed May 9, 2025

• Stable Virtual Camera

  #11151 closed May 8, 2025

• partial models spec in model_index.json

  #10413 closed May 8, 2025

• Question about non-convergence of training autoencoderkl.

  #11221 closed May 7, 2025

• ImportError: cannot import name 'CogView4LoraLoaderMixin' from 'diffusers.loaders'

  #11218 closed May 7, 2025

• Crash when loading Flux Schnell 1 model with train_dreambooth_lora_flux

  #11045 closed May 7, 2025

• Unable to load Flux LoRA trained with OneTrainer – NotImplementedError in `_convert_mixture_state_dict_to_diffusers`

  #11441 closed May 6, 2025

4 Issues opened by 4 people

• What's the difference between 'example/train_text_to_image_lora.py' and 'example/research_projects/lora/train_text_to_image_lora.py' ?

  #11542 opened May 13, 2025

• FluxPipeline produces noise when .enable_vae_slicing is used, and FluxImage2ImagePipeline does not support .enable_vae_slicing.

  #11540 opened May 12, 2025

• [Docs Update] AutoPipelineForInpainting.from_pretrained fails to load runwayml/stable-diffusion-inpainting without variant="fp16"

  #11528 opened May 9, 2025

• Request support for MAGI-1

  #11519 opened May 8, 2025

52 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [Quantization] Add TRT-ModelOpt as a Backend

  #11173 commented on May 12, 2025 • 2 new comments

• [LoRA] parse metadata from LoRA and save metadata

  #11324 commented on May 12, 2025 • 2 new comments

• docs: fix invalid links

  #11505 commented on May 10, 2025 • 1 new comment

• [Feature] AutoModel can load components using model_index.json

  #11401 commented on May 8, 2025 • 1 new comment

• SD3 ControlNet Script (and others?): dataset preprocessing cache depends on unrelated arguments

  #11497 commented on May 12, 2025 • 0 new comments

• hello i cant generate beautiful anime women

  #11345 commented on May 13, 2025 • 0 new comments

• [Community] Help us fix the LR schedulers when `num_train_epochs` is passed in a distributed training env

  #8384 commented on May 13, 2025 • 0 new comments

• PixArt Sigma PEFT LoRA loader support

  #11216 commented on May 13, 2025 • 0 new comments

• Improve `textual inversion` compatibility

  #10949 commented on May 9, 2025 • 0 new comments

• Fixes #11068 Add extended support for SD3 ControlNet (Stability AI)

  #11084 commented on May 11, 2025 • 0 new comments

• Bump jinja2 from 3.1.4 to 3.1.6 in /examples/server

  #11234 commented on May 8, 2025 • 0 new comments

• test new token for STYLE BOT

  #11257 commented on May 9, 2025 • 0 new comments

• Add VidTok AutoEncoders

  #11261 commented on May 9, 2025 • 0 new comments

• Fix the documentation of stable_diffusion_3.md

  #11287 commented on May 11, 2025 • 0 new comments

• [loaders] make the `loaders` folder structure cleaner

  #11335 commented on May 9, 2025 • 0 new comments

• Pipeline deprecations

  #11354 commented on May 6, 2025 • 0 new comments

• [chore] misc changes in the bnb tests for consistency.

  #11355 commented on May 9, 2025 • 0 new comments

• [LoRA training] add aspect ratio bucketing

  #11438 commented on May 8, 2025 • 0 new comments

• [NewModel] Adds SEVA ( Stable Virtual Camera) model

  #11440 commented on May 9, 2025 • 0 new comments

• [quant] add __repr__ for better printing of configs.

  #11452 commented on May 9, 2025 • 0 new comments

• [torch.compile] Make HiDream torch.compile ready

  #11477 commented on May 8, 2025 • 0 new comments

• [tests] Enable testing for HiDream transformer

  #11478 commented on May 7, 2025 • 0 new comments

• [tests] chore: rename lora model-level tests.

  #11481 commented on May 9, 2025 • 0 new comments

• [Modular diffusers] more refactors

  #11484 commented on May 13, 2025 • 0 new comments

• [feat] Add ostris / Flex.2-preview pipeline

  #11485 commented on May 6, 2025 • 0 new comments

• Change timestep device to cpu for xla

  #11501 commented on May 7, 2025 • 0 new comments

• AttributeError: _hf_hook caused by delattr in hooks.remove_hook_from_module()

  #10729 commented on May 6, 2025 • 0 new comments

• Error when I'm trying to train a Flux lora with train_dreambooth_lora_flux_advanced

  #11489 commented on May 6, 2025 • 0 new comments

• TypeError: UVit2DModel.forward.<locals>.layer_() got an unexpected keyword argument 'encoder_hidden_states'

  #11214 commented on May 6, 2025 • 0 new comments

• Sincerely Request The Support for Flux PAG Pipeline

  #11488 commented on May 6, 2025 • 0 new comments

• Expanded init fields in StableDiffusionPipeline cause incompatibilities with many/most inherited pipelines

  #6969 commented on May 7, 2025 • 0 new comments

• [Tests] fast tests single file checkpoints

  #6515 commented on May 8, 2025 • 0 new comments

• [Quantization] Support TRT as a backend

  #11032 commented on May 8, 2025 • 0 new comments

• [Models] handle initialization of new layers in a partially pre-trained model better

  #11279 commented on May 8, 2025 • 0 new comments

• [core] refactor `attention_processor.py` the easy way

  #10022 commented on May 8, 2025 • 0 new comments

• [tests] help us test `torch.compile()` for impactful models

  #11430 commented on May 8, 2025 • 0 new comments

• FlaxUNet2DConditionModel is not initialized with correct dtypes

  #11144 commented on May 8, 2025 • 0 new comments

• onnx export failure - timestep parameter with static value

  #11456 commented on May 8, 2025 • 0 new comments

• Add ReNeg: An end-toend method designed to learn improved Negative embeddings (CVPR 2025 Highlight)

  #11256 commented on May 9, 2025 • 0 new comments

• Error distributing FluxTransformer2DModel to multiple GPUs using controlnet

  #11247 commented on May 9, 2025 • 0 new comments

• Output of randn_tensor changes based on dtype, making seeds non-reproducible if the model precision is changed

  #11056 commented on May 9, 2025 • 0 new comments

• Support wan2.1 video model?

  #11008 commented on May 9, 2025 • 0 new comments

• Finetuned models aren't saved or loaded properly in train_custom_diffusion.py

  #9320 commented on May 10, 2025 • 0 new comments

• Full support for Flux attention masking

  #10194 commented on May 10, 2025 • 0 new comments

• [New Pipeline]: SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers

  #11135 commented on May 10, 2025 • 0 new comments

• InstructPix2Pix training script for SD3

  #9101 commented on May 11, 2025 • 0 new comments

• [Flux ControlNet] Add support for de-distilled models with CFG

  #9635 commented on May 11, 2025 • 0 new comments

• Error while loading Lora

  #11286 commented on May 11, 2025 • 0 new comments

• MultiControlNetModel is not supported for SD3ControlNetInpaintingPipeline

  #11208 commented on May 11, 2025 • 0 new comments

• HiDream auxiliary loss for MoE experts not tied to computation graph

  #11301 commented on May 12, 2025 • 0 new comments

• value errors in convert to/from diffusers from original stable diffusion

  #11285 commented on May 12, 2025 • 0 new comments

• Diffusers Transformer Pipeline Produces ComplexDouble Tensors on MPS, Causing Conversion Error

  #10986 commented on May 12, 2025 • 0 new comments