v2

I lowered the LR by increasing the resolution of the images in the dataset to improve the artifacts.

This LORA is still under-trained, in my opinion, but this method is too time consuming, so training was discontinued.

832 x 1216, 30 images.

No caption.

sd-scripts : fine tuning

optimizer : adafactor

scheduler : constant

LR : 2e-6

batch size : 6

epochs : 350

repeats : 1

v1

Slightly artifacted.

Enabled by prompts such as anime, manga, etc.

Trained on 416 x 608, 27 images.

No caption.

Parameters

sd-scripts : fine tuning

optimizer : adafactor

scheduler : cosine

LR : 7.5e-6

batch size : 3

epochs : 210

repeats : 1