A Lora single image training exercise, just for fun
Description
FAQ
Comments (6)
The title got me. I didn't see the meat in the background at first.
Brooo!, (or sis), you gotta teach us how to do lora training on a single image like that
1、prepare dataset like this video
https://www.youtube.com/watch?v=SSbrpAsnK3U
i suggest at least:30 clips
2、lora caption：Just use BLIP
3、USE Prodigy easy to train
4、save 1 /10 epochs, total epoch 100. total step must more than 1500.
5、num_dataset: if you just want get background. num>5 but epoch must down to 30(max step/(num*10epoch) ). if you want add some idea(like: 1girl) num=1 like this lora
6、batch size 1~2
dataset NUM
Panorama: Global Composition, =4 Middle ground: portrait shooting positioning Ratio =16 Close-ups: for details: =10
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.