Since the learning failed, I took revenge by lowering the LR and raising the epoch.
I'll report back below someday.
As a result, changing the LR does not improve learning.
https://civitai.com/articles/13094/optimal-design-for-background-loralora
Lowering the LR requires more steps to learn, but it doesn't dramatically improve background learning.β¬(In normal LR, learning is completed in 500steps.)
50 epoch,
Go ahead and upload yours!
Your query returned no results β please try removing some filters or trying a different term.