2 Releases published by 1 person

• v0.35.0 Diffusers 0.35.0: Qwen Image pipelines, Flux Kontext, Wan 2.2, and more

  published Aug 19, 2025

• v0.35.1 v0.35.1 for improvements in Qwen-Image Edit

  published Aug 20, 2025

26 Pull requests merged by 12 people

• continue translate document to zh

  #12194 merged Aug 20, 2025

• [chore] remove extra validation check in determine_device_map

  #12176 merged Aug 20, 2025

• Bria 3 2 pipeline

  #12010 merged Aug 20, 2025

• post release v0.35.0

  #12184 merged Aug 19, 2025

• translate document to zh

  #12179 merged Aug 19, 2025

• Performance Improve for Qwen Image Edit

  #12190 merged Aug 19, 2025

• Emergency fix for Qwen-Image-Edit

  #12188 merged Aug 19, 2025

• [Wan 2.2 LoRA] add support for 2nd transformer lora loading + wan 2.2 lightx2v lora

  #12074 merged Aug 19, 2025

• [chore] add lora button to qwenimage docs

  #12183 merged Aug 19, 2025

• Update README.md

  #12182 merged Aug 18, 2025

• [LoRA] feat: support more Qwen LoRAs from the community.

  #12170 merged Aug 18, 2025

• [docs] Clarify guidance scale in Qwen pipelines

  #12181 merged Aug 18, 2025

• fix(training_utils): wrap device in list for DiffusionPipeline

  #12178 merged Aug 18, 2025

• fix: caching allocator behaviour for quantization.

  #12172 merged Aug 18, 2025

• [Docs] typo error in qwen image

  #12144 merged Aug 18, 2025

• Minor modification to support DC-AE-turbo

  #12169 merged Aug 18, 2025

• [qwen] Qwen image edit followups

  #12166 merged Aug 18, 2025

• remove silu for CogView4

  #12150 merged Aug 18, 2025

• Qwen Image Edit Support

  #12164 merged Aug 17, 2025

• [docs] Quickstart

  #12128 merged Aug 15, 2025

• make test_gguf all pass on xpu

  #12158 merged Aug 15, 2025

• support hf_quantizer in cache warmup.

  #12043 merged Aug 14, 2025

• [core] respect local_files_only=True when using sharded checkpoints

  #12005 merged Aug 14, 2025

• feat: cuda device_map for pipelines.

  #12122 merged Aug 14, 2025

• [docs] Parallel loading of shards

  #12135 merged Aug 14, 2025

• make parallel loading flag a part of constants.

  #12137 merged Aug 14, 2025

18 Pull requests opened by 14 people

• [docs] Diffusion pipeline

  #12148 opened Aug 14, 2025

• Deprecate Flax support

  #12151 opened Aug 15, 2025

• [chore] remove controlnet implementations outside controlnet module.

  #12152 opened Aug 15, 2025

• `no_init_weights` in `from_pretrained` should be optional

  #12153 opened Aug 15, 2025

• [docs] AutoPipeline

  #12160 opened Aug 15, 2025

• make models and pipeline docs device agnostic, to cover accelerators other than CUDA

  #12161 opened Aug 15, 2025

• Propose to update SkyReels-V2

  #12167 opened Aug 17, 2025

• update document preview parts

  #12171 opened Aug 18, 2025

• [Modular] Provide option to disable custom code loading globally via env variable

  #12177 opened Aug 18, 2025

• feat(autoencoders): add patchify support for AutoencoderKLWan

  #12180 opened Aug 18, 2025

• Add device_map="auto" support for the diffusers pipelines with higher VRAM occupancy

  #12187 opened Aug 19, 2025

• [Wan 2.2 VAE] fix VAE tiling encode/decode

  #12191 opened Aug 19, 2025

• [scheduler] Add SDE Samplers: Curved Euler and Overshoot

  #12192 opened Aug 19, 2025

• Update README.md

  #12193 opened Aug 19, 2025

• [Modular] implement requirements validation for custom blocks

  #12196 opened Aug 20, 2025

• [Docs] Add documentation for KontextInpaintingPipeline

  #12197 opened Aug 20, 2025

• [docs] typo : corrected 'compile regions' to 'compile_regions'

  #12199 opened Aug 20, 2025

• [docs] Optimized code snippets

  #12200 opened Aug 20, 2025

11 Issues closed by 8 people

• Add support for guidance-interval during sampling

  #12198 closed Aug 20, 2025

• Wan2.2 Lightning LoRA Loading Fails

  #12147 closed Aug 19, 2025

• LoRA loading compatibility with some Qwen Image LoRAs

  #12159 closed Aug 18, 2025

• When training the Qwen-Image model using --bnb_quantization_config_path, the following error occurred: 'torch.device' object is not iterable.

  #12157 closed Aug 18, 2025

• Bad Image Quality generated by Qwen Image

  #12175 closed Aug 18, 2025

• "Requested to load AutoencoderKL"

  #12123 closed Aug 17, 2025

• IndexError: index 0 is out of bounds for dimension 0 with size 0

  #12104 closed Aug 17, 2025

• dje

  #12154 closed Aug 17, 2025

• cool man

  #12156 closed Aug 15, 2025

• monsieur

  #12155 closed Aug 15, 2025

• AttributeError: 'bool' object has no attribute '__module__'. Did you mean: '__mod__'?

  #12044 closed Aug 14, 2025

10 Issues opened by 10 people

• Version 0.35.0 AutoencoderKL failed with PyTorch 2.3.1: AutoAttributeError: module 'torch.library' has no attribute 'custom_op'

  #12195 opened Aug 20, 2025

• When using the qwen-image model to generate images, diffusers takes twice as long as diffsynth-studio

  #12189 opened Aug 19, 2025

• `WanVACETransformer3DModel` does not inherit `AttentionMixin` while `WanTransformer3DModel` does

  #12186 opened Aug 19, 2025

• What's the difference between DreamBooth LoRa and traditional LoRa?

  #12185 opened Aug 19, 2025

• 'super' object has no attribute '__getattr__'

  #12174 opened Aug 18, 2025

• Wan i2i with latent upscale

  #12173 opened Aug 18, 2025

• Failed to finetune the pre-trained model of 'stable-diffusion-v1-4' on image inpainting task

  #12165 opened Aug 17, 2025

• Not compatible with Flash ATTN 2.8.2

  #12162 opened Aug 16, 2025

• The diffusers0.34.0 installed by pip is not aligned with the source code

  #12149 opened Aug 15, 2025

• Wan 2.1 / Wan2.2 Lightning & Sparse Distillation & Tiny VAE Acceleration Methods

  #12146 opened Aug 14, 2025

22 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 Aug 20, 2025 • 9 new comments

• Fix AttributeError of `VisualClozeProcessor`

  #12121 commented on Aug 18, 2025 • 0 new comments

• [scheduler] Support custom sigmas/timesteps in UniPCMultistepScheduler

  #12109 commented on Aug 20, 2025 • 0 new comments

• [model] LTX Video 0.9.8

  #12095 commented on Aug 14, 2025 • 0 new comments

• Add Dream 7B Diffusion Large Language Model Pipeline

  #12091 commented on Aug 20, 2025 • 0 new comments

• gguf new quant type support (with demo)

  #12076 commented on Aug 18, 2025 • 0 new comments

• [lora] factor out the overlaps in `save_lora_weights()`.

  #12027 commented on Aug 20, 2025 • 0 new comments

• Fixed "NameError: name 'ftfy' is not defined" in Wan pipeline by directly importing ftfy.

  #12023 commented on Aug 16, 2025 • 0 new comments

• Add F5 TTS pipeline

  #11958 commented on Aug 16, 2025 • 0 new comments

• Context Parallel w/ Ring & Ulysses & Unified Attention

  #11941 commented on Aug 14, 2025 • 0 new comments

• Propose to refactor `norm_out` in `SkyReelsV2Transformer3DModel` for device handling in multi-GPU case

  #11850 commented on Aug 20, 2025 • 0 new comments

• Add MAGI-1: Autoregressive Video Generation at Scale

  #11713 commented on Aug 20, 2025 • 0 new comments

• [bugfix] Fix the issue that will save the model card if push to hub is not applied in HiDream

  #11629 commented on Aug 14, 2025 • 0 new comments

• Support for control-lora

  #10686 commented on Aug 20, 2025 • 0 new comments

• How to finetune Flux-dev full params, 80G OOM ...

  #10196 commented on Aug 20, 2025 • 0 new comments

• train_dreambooth_lora_flux_kontext.py: error: the following arguments are required: --instance_prompt

  #11973 commented on Aug 20, 2025 • 0 new comments

• inference with lora suddenly significantly slower

  #12131 commented on Aug 19, 2025 • 0 new comments

• Will `cosmos-transfer1` be supported in diffusers in the future?

  #12084 commented on Aug 19, 2025 • 0 new comments

• requirest for the model use

  #12138 commented on Aug 17, 2025 • 0 new comments

• The IP Adapter provides different conditional guidance for each frame generated by AnimateDiff.

  #12139 commented on Aug 17, 2025 • 0 new comments

• Error creating pre-trained FluxPipeline with diffusers "Cannot instantiate this tokenizer from a slow version"

  #9063 commented on Aug 16, 2025 • 0 new comments

• resize_token_embeddings (?) (different result from new and old version)

  #12145 commented on Aug 14, 2025 • 0 new comments