Load Flux2/Klein LoRAs as native NetworkModuleLora objects, bypassing
diffusers PEFT. Handles kohya (lora_unet_), AI toolkit (diffusion_model.),
diffusers PEFT (transformer.), and bare BFL key formats with automatic
QKV splitting for double block fused attention weights.
Includes shape validation to reject architecture-mismatched LoRAs early.
Respects lora_force_diffusers setting to fall back to PEFT when needed.
- detect f2 model type for LoRAs via metadata, architecture, and filename/folder
- preprocess bare BFL-format keys with diffusion_model prefix for Flux2LoraLoaderMixin
- handle LoKR format via native NetworkModuleLokr with on-the-fly kron(w1, w2)
- add NetworkModuleLokrChunk for fused QKV split into separate Q/K/V modules
- activate native modules loaded via diffusers path
- improve error message for Flux1/Flux2 architecture mismatch
Add 4-step distilled Nunchaku SVDQuant entries for Qwen-Lightning and
Qwen-Lightning-Edit alongside the existing 8-step variants. Step count
is now shown in the reference name (e.g. "Qwen-Lightning (4-step)").
- Add subfolder parameter to load_qwen_nunchaku to distinguish
4-step (nunchaku-4step) from 8-step (nunchaku) variants
- Route to correct safetensors: lightningv1.0-4steps vs
lightningv1.1-8steps for gen, lightningv1.0-4steps vs
lightningv1.0-8steps for edit
- Strip nunchaku subfolder before pipeline from_pretrained since
it does not exist in the base HuggingFace repos
- Rename HuggingFace org from nunchaku-tech to nunchaku-ai across all
nunchaku model repos (flux, sdxl, sana, z-image, qwen, t5)
- Add per-torch-version nunchaku version mapping instead of single global
version, with robust torch version parsing
- Relax sd_detect to match 'anima' without requiring 'cosmos' in name
- Use hf_hub_download for custom pipeline.py and adapter modules
- Register custom modules in sys.modules for Diffusers trust_remote_code
- Pass trust_remote_code=True to from_pretrained
- Map AnimaTextToImage to 'cosmos' model type for TAESD preview support
Anima replaces the Cosmos T5-11B text encoder with Qwen3-0.6B + a
6-layer LLM adapter and uses CONST preconditioning instead of EDM.
- Add pipelines/model_anima.py loader with dynamic import of custom
AnimaTextToImagePipeline and AnimaLLMAdapter from model repo
- Register 'Anima' pipeline in shared_items.py
- Add name-based detection in sd_detect.py
- Fix list-format _class_name handling in guess_by_diffusers()
- Wire loader in sd_models.py load_diffuser_force()
- Skip noise_pred callback injection for Anima (uses velocity instead)
- Add output_type='np' override in processing_args.py
- Add Transformers v5 tokenizer compatibility fix for SDNQ Klein models
Downloads missing vocab.json, merges.txt, tokenizer_config.json from
Z-Image-Turbo when needed
- Detect SDNQ repos and disable shared text encoder to use pre-quantized
weights from the SDNQ repo instead of loading from shared base model
- Update reference-quant.json with correct preview images and metadata
for Klein SDNQ models
- Update reference-distilled.json with correct cfg_scale (1.0) for
distilled Klein models per official HuggingFace documentation
- Add 6 Klein model preview images
- Fix Klein text encoder comment to specify correct sizes per variant
- Lock TAESD decode logging behind SD_PREVIEW_DEBUG env var
- Fix misleading comment about FLUX.2 128-channel reshape (is fallback)
- Remove VRAM requirements from model descriptions in reference files
Add support for FLUX.2 Klein distilled models (4B and 9B variants):
- Add pipeline loader for Flux2KleinPipeline
- Add model detection for 'flux.2' + 'klein' patterns
- Add pipeline mapping in shared_items
- Add shared Qwen3ForCausalLM text encoder handling:
- 4B variants use Z-Image-Turbo's Qwen3-8B
- 9B variants use FLUX.2-klein-9B's Qwen3-14B
- Add reference entries for distilled (4B, 9B) and base models
- Update diffusers commit for Flux2KleinPipeline support
- Add GLM-Image (zai-org/GLM-Image) model detection and loading
- Custom pipeline loader with proper component handling:
- ByT5 text encoder (cannot use shared T5 due to different hidden size)
- Vision-language encoder (9B AR model)
- DiT transformer (7B)
- Fix EOS token early stopping in AR generation
- Add AR token generation progress tracking with terminal progress bar
- Fix uninitialized audio variable in processing
- Add TAESD support for GLM-Image (using f1 variant)
awsr
vladmandic
CalamitousFelicitousness
Oleksandr Liutyi