
Honestly, I'm not happy with this model ?. Can anyone give me tips on how to improve it? ? Here are the parameters: https://civitai.com/articles/5255/my-sdxl-lora-experience-12gb-rtx3060-gpu-local-training-wkohya-ss
Steps - 205 per epoch
Unique images - 75
Dir Structure
-
1_close-Up - 40 Images
-
1_undercarriage - 45 Images (in which 30 Images is from 1_close-Up + 15 Images is from 2_flying )
-
2_dragon-rider - 30 Images (in which 15 Images is from 2_flying )
-
2_flying - 30
(It seems I forgot to add the class prompt ?.)"
Total epochs - 40
This file only include 39th epoch
Working properly (Trigger word combination) :-
-
dragon-rider
-
close-up
-
flying
-
undercarriage
-
undercarriage, close-up, flying
-
undercarriage, close-up
Not Working properly (Trigger word combination) :-
-
dragon-rider, flying (some time woks)
-
undercarriage, dragon-rider, flying (only 2 images exist of this)
-
dragon-rider, close-up
-
(do not work properly with any character Lora)
描述:
训练词语: close-Up,undercarriage,dragon-rider,flying
名称: UDS_Concept_DragonRider.safetensors
大小 (KB): 223100
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success