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/regional-conditioning-tuning 分支b5c334d8ca Fix _negative_cross_attn_mask_score.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/regional-conditioning-latent-area 分支比较 2 提交 »ef8177ba4b Get area-based regional prompting working.bc869cad62 First draft of area-based latent-space regional prompting.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/regional-conditioning-cross-attn-mask-only 分支比较 10 提交 »61cd889dc5 Remove dispatch_progress() function that was added aciidentally during conflict resolution.d64df4a260 Remove AddConditioningMaskInvocaton.644232f7ea Default model settings (#5850)e2448d2b4e Next: Allow in place local installs of models60815f59d1 Log a stack trace for invocation errors.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/raw-model-circular-import 分支ed5b106ded Empty out invokeai/backend/model_manager/__init__.py and use full absolute imports instead.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/prompt-augment-experiments 分支比较 4 提交 »8c09b345ec wipb7a1086325 Attempt at style prompt that did not work very well.3062fe2752 Add phi-3 test script.76a65d30cd Naive prompt augmentation experiment with GPT-2 - results not great yet.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/peft-3 分支比较 2 提交 »bfd5cdb311 (minor) tidy types in sdxl_format_utils.py24950dea8c Copy existing utils for converting SDXL LoRA state dict formats to a new file.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/peft-2 分支比较 2 提交 »f9fda503a3 A very primitive working version of peft patching. It is very slow. LoRAs don't get unloaded yet, so can only be run once. And the results are *slightly* different than the old implementation. I suspect this is because the lora weight is not being applied to the UNet, but there could be other issues as well.22c66cf55b WIP