huggingface/diffusers

View on GitHub

🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.

Stars: 34.5k
Forks: 7.3k
Open beginner issues: 22
Indexed issues: 838
Avg merge: 3d 3h
Merged PRs (30d): 91

Dominant language: Python
License: Apache-2.0
Last GitHub push: Sep 15, 2026
Latest indexed: Sep 17, 2026
Contributing guide: Contributing guide
Code of conduct: Code of conduct
Beginner labels: help wanted good first issue

Open All

838 open issues indexed

[meta issue] Systematic model/pipeline review findings / tracking Open

huggingface/diffusers#13656 · 11 comments · 1 reaction · 0 assignees · Indexed Sep 17, 2026
如何实现flux2Klein的DPO训练 Open

huggingface/diffusers#13682 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
universal method or class to load any model locally Open

feature-request gguf pipelines quantization single-file

huggingface/diffusers#13683 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
FluxKlein Training Scripts - CFG issue Open

huggingface/diffusers#13762 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Bad image output for Flux.2-dev, rocm, quantization and separate prompt encoding sequence Open

bitsandbytes bug models pipelines quantization

huggingface/diffusers#13772 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
UNet2DModel dtype property fails under nn.DataParallel with UnboundLocalError in get_parameter_dtype Open

bug models

huggingface/diffusers#13789 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
loading Cosmos2 pipeline also disabled gradient tracking globally Open

bug pipelines

huggingface/diffusers#13790 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Incompatibility between FlashAttention and ERNIE-Image Open

bug pipelines

huggingface/diffusers#13801 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Fix incorrect batch handling in _prepare_image_ids usage in train_dreambooth_lora_flux2_img2img.py Open

bug

huggingface/diffusers#13811 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
[Community Support] Integrating visual generative foundation models in diffusers Open

huggingface/diffusers#13844 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
fix underlying issue with `test_from_save_pretrained_dtype_inference` is that the `model.to(dtype)` cast at Open

huggingface/diffusers#13869 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
When use UniPCMultistepScheduler as the ODE solver, and input different num_inference_steps in StableDiffusionXLPipeline, the first inference step is wrong Open

bug needs-code-example pipelines schedulers

huggingface/diffusers#13887 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
[New Pipeline/Model] Add JoyAI-Echo multi-shot audio-video generation pipeline Open

feature-request models pipelines

huggingface/diffusers#13909 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
1 bit, ternary image generation support Open

feature-request pipelines quantization

huggingface/diffusers#13925 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Please reconsider `transformers>=5.0.0` Open

bug

huggingface/diffusers#13935 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
FLUX LoRA with CLIP text-encoder weights fails (empty rank -> IndexError) under transformers>=5 Open

bug lora pipelines

huggingface/diffusers#13984 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Support Original Checkpoint-Compatible PEFT Adapters Open

huggingface/diffusers#14002 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Split Q,K,V vs Fused QKV Projections in diffusers Discussion Open

huggingface/diffusers#14003 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
PixArtSigmaPipeline: no LoRA loading support Open

huggingface/diffusers#14004 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
from_pretrained fails loading deprecated pipelines Open

huggingface/diffusers#14008 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
attention dispatcher assumes wrong attributes for flash attn kernel from hub Open

bug

huggingface/diffusers#14012 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
MagCache on Wan 2.2 Dual-Transformer Pipelines: Incorrect Step Accounting and Limited Effectiveness on a 4-Step Distilled Model (e.g. Wan2.2) Open

huggingface/diffusers#14025 · 3 comments · 1 reaction · 0 assignees · Indexed Sep 17, 2026
Remove link to HEIM for diffusion model evaluation Open

huggingface/diffusers#14033 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Finalize Guidelines for Ulysses based context parallel attention Open

huggingface/diffusers#14041 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
SD1.5 - pipeline_controlnet_img2img and pipeline_controlnet_inpaint are mixing variables "image" and "control_image" Open

bug

huggingface/diffusers#14057 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Backport workflow-hardening fix (excessive-permissions) to 3 release branches Open

huggingface/diffusers#14085 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Add SeFi-Image pipeline Open

huggingface/diffusers#14086 · 0 comments · 1 reaction · 0 assignees · Indexed Sep 17, 2026
Handling of .alpha in from-comfy converters Open

bug

huggingface/diffusers#14088 · 8 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Regression (#13485) Broken TorchAO Compat Open

bug

huggingface/diffusers#14097 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Flux-family attention: flash-attn and other backends fail in an autocast context Open

bug

huggingface/diffusers#14104 · 6 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
`from_single_file()` support missing/incomplete for Ideogram4 and Krea2 transformers Open

bug feature-request models single-file

huggingface/diffusers#14122 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
[Version 0.39.0]AttributeError: 'CLIPTextModel' object has no attribute 'text_model' WHEN SET CKIP_SKIP NOT NONE Open

bug pipelines

huggingface/diffusers#14124 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Boogu-Image Open

feature-request models

huggingface/diffusers#14138 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Non-hub varlen attention backends mispack keys for non-contiguous masks (missed in #14114/#14115) Open

bug models

huggingface/diffusers#14139 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
DiffusionPipeline.download() on Windows silently skips all subfolder config.json files with huggingface_hub>=1.22.0 Open

bug pipelines

huggingface/diffusers#14142 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
`LTX2*Pipeline.from_single_file` cannot load LTX 2.3 (or 2.0) combined checkpoints: unmapped 2.3 keys leave meta tensors, and the LTX2 converters drain the shared checkpoint dict Open

bug models pipelines single-file

huggingface/diffusers#14145 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Add AMD Quark support to Diffusers Open

feature-request models quantization

huggingface/diffusers#14149 · 1 comment · 1 reaction · 0 assignees · Indexed Sep 17, 2026
[Bug] QwenImageEdit produces incorrect outputs with transformers >=5 due to missing mm_token_type_ids Open

bug pipelines

huggingface/diffusers#14194 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
LDM3D rgblike_to_depthmap truncates the 16-bit depth map back to the 8-bit input dtype Open

bug pipelines

huggingface/diffusers#14206 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
VaeImageProcessor.blur() docstring is missing blur_factor and has a copy-pasted description Open

bug needs-code-example pipelines

huggingface/diffusers#14209 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Ancestral schedulers (EulerAncestral, KDPM2Ancestral) produce all-NaN output with beta_schedule="squaredcos_cap_v2" Open

bug schedulers

huggingface/diffusers#14213 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
LTX2Pipeline: dynamic timestep shift `mu` is constant, ignores the actual sequence length Open

bug pipelines

huggingface/diffusers#14243 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
PR GPU CI skips new-model GPU tests: Cosmos3 memory tests failed on an unrelated PR Open

bug models needs-code-example

huggingface/diffusers#14247 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
`set_attention_backend` on one model leaks globally to other models via the active-backend registry Open

huggingface/diffusers#14249 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
[tests] migrate all attention processor related stuff from pipeline to model-level testing Open

huggingface/diffusers#14255 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
Add `diffusers-cli` support for quantization via `PipelineQuantizationConfig` Open

huggingface/diffusers#14259 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
Mage-Flow Open

huggingface/diffusers#14270 · 0 comments · 3 reactions · 0 assignees · Indexed Sep 17, 2026
[Cache] Add First Block Cache (FBCache) support for Flux 2 (Flux2Transformer2DModel) Open

huggingface/diffusers#14280 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Skill versioning & update strategy for `diffusers-cli skills` Open

huggingface/diffusers#14281 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Modular] Krea 2 image-to-image generation Open

huggingface/diffusers#14290 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026

huggingface/diffusers

Get new issues in your inbox