Lora Training

After training a LoRA (for the first time) using the website it did not go great my dataset and caption were that great to begin with (my fault totally). However, I am curious about what this Training Attempts 1 / 3 means (it is on the preview page). Can I "retrain" the LoRA?

I am fraid you can not.

Because retrain is actually another time to run the same python and use GPU to really [hot work].

But here are some suggestions I would like to share with you.

1. More is less than Good. Choose some really Neat Datebase Pictures, and descripe the things is common but you [do not want] into caption.

2. Reaction to images and gain Buzzs 100/day. and try to do some bounty to win the reward.

3. Training Attempts means that the detailed parameter settings are different, such as attenuation ratio, noise ratio, etc. But don't worry about it they would not influence too much.