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-lora-diffusers_v2 分支比较 8 提交 »766c6082a3 Add state dict keys for a new FLUX LoRA diffusers variant.ad5d528204 Fix frontend lint errors.536ccf071c (minor) remove remaining incorrect references to T5 encoder in comments.ca13c3b12f Update Linear UI to support FLUX LoRA models that patch the CLIP model in addition to the transformer.d54c1ef9ba Whoops, the 'lora_te1' prefix in FLUX kohya models refers to the CLIP text encoder - not the T5 as previously assumed. Update everything accordingly.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-lora 分支比较 10 提交 »0ff5355ce3 Rename flux_kohya_lora_conversion_utils.py3ad5fc060d Fixup FLUX LoRA unit tests.17d5c85454 WIP4698649cc9 WIP - add invocations to support FLUX LORAs.d41c075768 Get probing of FLUX LoRA kohya models working.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-ip-adapter-cfg 分支比较 3 提交 »125be54b95 Use a black image as the negative IP prompt for parity with X-Labs implementation.b74dcfcef0 Test out IP-Adapter with CFG.494cedf090 Merge branch 'ryan/flux-cfg' into ryan/flux-ip-adapter-cfg
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-inpainting-change-schedule 分支比较 6 提交 »cd931d5a1c Best heuristic so far.6d1aeb94f0 WIP - tuning noise guidance paramsf4a1cbf8ad Image to image noise guidance methods.c67a4277c9 Linear inpaint change schedule.d96ca6bd79 Interesting checkpoint - change val to see effect.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-inpainting 分支c53ec3de45 checkpoint
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-improved-tg-3 分支比较 2 提交 »d122343431 Disable TGS in frontend for testing.0139788e65 Fractional denoise start.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-improved-tg-2 分支比较 2 提交 »7e6dbba8ca Apply to all modes in the frontend.e11f98f128 experiment 2
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-fast 分支比较 4 提交 »182c5793ba Commented pytorch profiler.675a66612c Avoid unnecessary dtype conversions with rope encodings.abdf2a7f86 Use view() instead of rearrange() for better performance.bb098ec064 Replace custom RMSNorm implementation with torch.nn.functional.rms_norm(...) for improved speed.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/flux-dynamic-seq-len 分支比较 2 提交 »8d04ec3f95 Improve docs related to dynamic T5 sequence length selection.4581a37a48 Dynamically select smaller t5 seq len to save inference time.