12月1月2月3月4月5月6月7月8月9月10月11月
周一周三周五
一年内 265876 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-dora-one-trainer 分支比较 10 提交 »67afa7e339 Update PartialLayer to work with unquantized / GGML quantized / BnB quantized layers.92c6a7d658 Add FLUX OneTrainer model probing.caa9ecafae Finish switching LoRAModelRaw to use a list of names/layers rather than a dict. This enables multiple entries with the same name.f21464e972 Relax lora_layers_from_flux_diffusers_grouped_state_dict(...) so that it can work with more LoRA variants (e.g. hada)eeff9d3df5 Fix bug in FLUX T5 Koyha-style LoRA key parsing.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-dora 分支0fb3b53446 Add test state dict for a FLUX DoRA model.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-controlnet-xlabs 分支比较 10 提交 »a7ded2e1f8 Add support for FLUX controlnet weight, begin_step_percent and end_step_percent.acfed395fb First pass at integrating FLUX ControlNets into the FLUX Denoise invocation.cf32c7b370 Add FLUX XLabs ControlNet model probing.9d897b46cb Fix type errors and imporve docs for ControlNetFlux.9ad64bb45b Remove gradient checkpointing from ControlNetFlux.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-2 分支比较 9 提交 »fb5db32bb0 Make it run (with artifacts).823c663e1b WIP on moving from diffusers to FLUXd40c9ff60a More improvements for LLM.int8() - not fully tested.373b46867a LLM.int8() quantization is working, but still some rough edges to solve.dc66952491 Clean up NF4 implementation.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flex-1-alpha 分支231099f913 Infer FLUX model params from the shape of the state dict rather than assuming that the model is either Dev or Schnell. This enables the FLEX.1-alpha model.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/fix-flux-lora-no-proj-mlp-2 分支比较 3 提交 »22c53dce45 Add unit tests for FLUX diffusers LoRA without .proj_mlp layers.ba3cd9877c Add state dict tensor shapes for existing LoRA unit tests.05f3492e38 Support FLUX LoRA models without .proj_mlp layers.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/fix-depth-anything-postprocessing 分支e22f0f2203 Update DepthAnything post-processing logic to avoid artifacts caused by numerical overflow.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/fix-calc-model-size 分支比较 5 提交 »0781fdf3b0 WIP - simplify ModelLoadRegistry8d7ca9c1b7 More refactoring to help with circular imports.a61e0bd2dd Remove symbol re-exports that were contributing to circular import issues.798e73969c Tidy handling of SCHEDULER_NAME_VALUES to help with circular import errors.44f62944ee Fix circular import caused by the organization the model size utils.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/fix-anti-patterns 分支比较 2 提交 »3f2dedceec Remove unused check_internet() function.0bdf7f5726 Do less stuff on import of api_app.py. Instead, call functions imperatively.