Trained and only for tests.
All training pictures are from the internet.
Do not use this model in commercials!
v1.5: Removed the watermark!
Description
FAQ
Comments (18)
这扳机词可真壮观。。。
大佬,请教一下,为啥我试了一下训练lora,出来才9M大小,您这是多少照片训练出来的,,怎么这么大。。
如果你用的是sd-scripts,请将 network_dim 这个参数改成128即可获得大约144MB的lora,但这个参数并不是越大越好,32或者64同样也可以训练出很多出色的lora
AchilliceMar 5, 2023
@zyx123 請問能分享一下怎麼訓練的嗎?
網上教的一堆都是假的..練不出來..
@empressa7753133 你可以试试这个 GitHub - Akegarasu/lora-scripts: LoRA training scripts use kohya-ss's trainer, for diffusion model.
那个,请问用了多少张图,一张图大概学习多少次比较好呀?
大约150张图,学习了20000+次,这方面我也不太懂,仅作参考
5201linyi925Apr 5, 2023
@zyx123 每张图都要白色背景吗,她图上的水印要去掉吗,我的脸练的走样了,不像她,而且头还有点大
@5201linyi925 不需要白色背景,推荐去水印(可以用lama-cleaner)
人类的英雄
为什么我画出来像个战鹰qwq
非常喜欢这个妹子 可惜是1.5的
Details
Downloads
12,239
Platform
CivitAI
Platform Status
Available
Created
3/4/2023
Updated
8/25/2026
Deleted
-
Trigger Words:
1girl
solo
lips
looking at viewer
realistic
breasts
underwear
long hair
brown eyes
panties
black hair
sitting
bra
brown hair
medium breasts
cleavage
jewelry
bare shoulders
hair ornament
twintails
short hair
earrings
school uniform
serafuku
foot focus
twin braids
china dress
chinese clothes
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.

