Lora trained on 218 images of Quinella.
Description
FAQ
Comments (2)
hello, very good, your loRA wanted to know how you do it, do you have a link where you learned from, thank you very much
Hi there
I use this:
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts/releases/tag/installers-v5
I use default settings mostly.
I have a 3070ti so I do batch size 2, approx 5 repeats depends on number of images in dataset, 40 epochs, 8 dim, cosine with restarts, I prune data set tags when making characters and sex act loras using https://github.com/toshiaki1729/stable-diffusion-webui-dataset-tag-editor
and that's about it.
If you prune tags for character and want a custom prompt/tag that you add yourself, you normally want to keep n tags at front and shuffle.
I normally just delete bad art from the dataset \o/
I use grabber to just rip images and their tags from gelbooru for my datasets.