
V2.0 is an uncompleted Lora due to the support issues of WebUI to SDXL models.
I hope someone can help me out.(questions post below)
What's New?
-
196 Highres Pictures dataset.(Compared to 70 pics for V1.0)
2.AI enhanced Quality and Aesthetic.(More brutal and monumental)
Problems...
-Couldn't merge Lora with block weight information in Plug-in called Supermerger.
And I got this error:
-SDXL model couldn't work with tile diffusion for highres fix in imag2imag
Progress:
Lora weight should not exceed 0.7
Lora Block Weight has been tested and adjustment was ready, unfortunately couldn't correctly bake information into this Lora:
y-axis is Lora weight for the one I want to mix with. (my Brutalism V1.0)
So what you expect will be the "ALL" performance instead of the "TEST" (which should be):
I really appreciate If someone can help me out of those problems,
And you are welcome to feed your creative work back!
Thank you.
描述:
This version is a realistic-based LoRa, and it's trained for Photo-realistic architecture generation. My goal is to create a SDXL LoRa which can produce an architecture style combining brutalism and Sci-Fi elements (like those in Death Stranding).
训练词语: Soviet,building,scenery
名称: Soviet-010.safetensors
大小 (KB): 445787
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success