Issues

• Kandinsky 6: add context-parallel (_cp_plan) support to Kandinsky6Transformer3DModel

  huggingface/diffusers#15156 · Filter by author vladislav-vasilenko opened on Oct 30, 2026
  1

• Clarify set_adapters() adapter_weights documentation and type hints

  huggingface/diffusers#15146 · Filter by author gorkadabo1 opened on Oct 29, 2026
  1

  

• flux_lora_quantization example: compute_embeddings.py crashes on current transformers

  bug
  huggingface/diffusers#15145 · Filter by author kahlun opened on Oct 29, 2026

• DDPM schedulers add fresh noise on terminal transitions when the final timestep is nonzero

  huggingface/diffusers#15139 · Filter by author surgan12 opened on Oct 27, 2026
  3

• [Feature Request] Add DiCache for sample-adaptive caching in FLUX transformers

  huggingface/diffusers#15134 · Filter by author BlazeSR10hawk opened on Oct 26, 2026
  1

• [Feature] Reuse Ulysses shard lengths across MiniMax-H3 blocks

  huggingface/diffusers#15132 · Filter by author LeslieWylie opened on Oct 26, 2026

• Flux2KleinInpaintPipeline: padding_mask_crop pastes a center crop of the result instead of resizing it

  bug
  huggingface/diffusers#15131 · Filter by author hkang4628 opened on Oct 26, 2026
  4

• @dataclass on quant configs breaks __eq__/__repr__/hashing, and to_json_string() raises for most configs

  huggingface/diffusers#15126 · Filter by author OnePunchMonk opened on Oct 24, 2026

• Nunchaku AWQ config accepts any group_size, but the kernel only supports 64

  huggingface/diffusers#15125 · Filter by author OnePunchMonk opened on Oct 24, 2026

• Quantizer configs are mutated in place, leaking modules_to_not_convert across loads

  huggingface/diffusers#15124 · Filter by author OnePunchMonk opened on Oct 24, 2026
  3

• from_single_file forces pre_quantized=True, breaking bnb/TorchAO/Quanto single-file loading

  huggingface/diffusers#15123 · Filter by author OnePunchMonk opened on Oct 24, 2026

• NVIDIAModelOptConfig.to_dict() crashes from_pretrained/save_pretrained when forward_loop is set

  huggingface/diffusers#15122 · Filter by author OnePunchMonk opened on Oct 24, 2026

• AutoRound: user-passed quantization_config silently overwrites checkpoint's quant settings

  huggingface/diffusers#15120 · Filter by author OnePunchMonk opened on Oct 24, 2026

• MiniMax-H3 creates full-clip FP32 temporaries after VAE decoding

  huggingface/diffusers#15119 · Filter by author LeslieWylie opened on Oct 24, 2026

• [Model integration] Add Hy-Nomas support to Diffusers

  huggingface/diffusers#15113 · Filter by author jiaxiangc opened on Oct 23, 2026
  1

• Text-to-image example updates EMA when AMP skips the optimizer step

  huggingface/diffusers#15112 · Filter by author SeverinVisionary opened on Oct 23, 2026
  1

• LMSDiscreteScheduler uses the wrong multistep order when sampling starts part-way through the schedule (img2img, inpainting)

  huggingface/diffusers#15108 · Filter by author mottopanikeiku opened on Oct 22, 2026
  1

• encode_video truncates a fractional fps, so the video runs longer than its audio

  huggingface/diffusers#15104 · Filter by author devanshc777 opened on Oct 22, 2026
  1

• [modular] ConditionalPipelineBlocks should declare its own trigger inputs

  huggingface/diffusers#15101 · Filter by author yiyixuxu opened on Oct 22, 2026
  3

• [modular] conditional block triggers should be workflow inputs

  huggingface/diffusers#15100 · Filter by author yiyixuxu opened on Oct 22, 2026

• load_video returns mixed-mode frames for GIFs, breaking preprocess_video and AnimateDiff video-to-video

  huggingface/diffusers#15098 · Filter by author nettanvirdev opened on Oct 21, 2026

• DEIS coefficient precision causes increasing error as inference steps increase

  huggingface/diffusers#15090 · Filter by author Robertboy18 opened on Oct 21, 2026
  1

• lr_scheduler step bug in examples code

  bug
  huggingface/diffusers#15087 · Filter by author weixiaohao opened on Oct 20, 2026
  2

• Qwen-Image 2.1 Turbo: VAE tiling introduces vertical streaks at 2048x2048 on NVIDIA/CUDA

  huggingface/diffusers#15083 · Filter by author Religious-J opened on Oct 19, 2026
  4

• Follow-up to #14737: drop output_type="latent" from modular decode blocks

  huggingface/diffusers#15071 · Filter by author yiyixuxu opened on Oct 17, 2026
  2