Finally save enough buzz for my first lora training 0w0
This model is to reproduce one of my favorite artists -- kurokawasudo's style, especially for the muscle and facial expression.
Since this is my first time training a Lora, I am not sure if my parameter adjustments are appropriate. I would greatly appreciate any suggestions.
I am saving up for buzz to train my next Lora, hope next one could be better!
Don't forget to support original artist: sudo (@kurokawasudo) / X
Trigger word: sud0_test01
Description
FAQ
Comments (3)
👏👏👏 Thanks!
Any recommendations for the steps and CFG scale? Are there specific trigger words?
Oh! I forgot to include the trigger word in the description — just updated it. Please use sud0_test01 as the trigger word.
The number of steps and CFG scale mostly depend on the checkpoint you're using. In general, I use around 20–25 steps and CFG scale of 5–7.
@kiork thanks!