All pull requests

Pull requests

• fix: ComfyUI modern API compatibility, Windows HostBuffer crash prevention, and DiT model caching

  #829·LacklusterOpsec opened Aug 26, 2026·

  ·

• fix: weight dereferenced before torch_dtype/device kwargs are honoured (ComfyUI 0.33 lazy Linear)

  #828·JoaoZaokk opened Aug 18, 2026·

  ·

• feat: fix Windows AIMDO crash, add Qwen quantized text encoders & FLUX.2 Klein

  #825·Sniper199999 opened Jul 10, 2026·

  ·

• Support lazy-init Linear layers in Z-Image patching

  #817·kj863257 opened Apr 7, 2026·

  ·

• fix: match rope batch dim with input

  #802·ssitu opened Feb 3, 2026·

  ·
  1

  3

• Update setup_custom_nodes.py

  #799·wzgrx opened Feb 1, 2026·

  ·
  1

• feat: enable dynamic cpu-offloading for qwen image based models

  #791·Cyclica-bin opened Jan 25, 2026·

  ·

• feat: add portable windows build script and documentation

  #786·Bluear7878 opened Jan 19, 2026Contributor·

  ·
  1

• Fix crash in patch_scale_key when tensor dtypes do not match and prevent prompt corruption

  #754·1038lab opened Dec 30, 2025·

  ·
  6

• feat: add LoRA support for Z-Image turbo models

  #739·phatwila opened Dec 26, 2025·

  ·
  17

• 1.Fix flux lora leakage 2. Fix Qwen image sample bug for new comfyui version (v0.6.0) . 3. Fix Qwen image lora 3-party code for new comfyui version

  #732·GuardSkill opened Dec 24, 2025·

  ·

• Fix QwenImage _forward argument alignment for ComfyUI master（修复 QwenImage 在 ComfyUI master 下的参数错位问题）

  #728·VipCrazyMaster opened Dec 22, 2025·

  ·
  10

• fix NunchakuFluxPuLIDApplyV2 bug in new version comfyui [Bug] TypeError: 'NoneType' object is not callable in pulid.py #719

  #720·Susiia opened Dec 12, 2025·

  ·
  4

• fix: prevent LoRA weight accumulation in Flux loader

  #718·Cor-net opened Dec 11, 2025·

  ·
  1

• feat: Pi-Flow (low step adapters) for FLUX-dev

  #697·Pizzawookiee opened Nov 15, 2025·

  ·
  1

• fix: ensure contiguous tensors after slicing in QwenImage attention

  #696·diodiogod opened Nov 13, 2025·

  ·
  1

• Apply x_embedder to controlnet to match hidden_states shape

  #691·chriscombs opened Nov 8, 2025·

  ·
  1

• SDXL + SDXL Turbo UNetLoader, LoRA, FB Cache, ControlNet support

  #680·Pizzawookiee opened Oct 27, 2025·

  ·
  2

• Update requirements.txt

  #674·wzgrx opened Oct 25, 2025·

  ·
  1

• feat: Update qwen to support in loras version 1.0.0 (standalone)

  #665·GavChap opened Oct 17, 2025·

  ·
  34

• feat: Add support for Qwen Image LoRA

  #642·Bluear7878 opened Oct 7, 2025Contributor·

  ·
  1

  4

• Fix CUDA kernel error on Turing GPUs when resetting LoRA (#418)

  #629·lvyonghuan opened Oct 2, 2025·

  ·
  1

• Fix crash when broken symlinks exist in diffusion_models folder

  #628·diodiogod opened Oct 2, 2025·

  ·