When I trained lora model based playground v2.5, it didn't work very well #7624
Unanswered
plastic0313
asked this question in
Q&A
Replies: 2 comments 3 replies
-
The dynamics of EDM training is not fully well-known to us. You should run experiments with different hyperparameters and see what works best. The values provided in the official example lead to reasonable results, though. |
Beta Was this translation helpful? Give feedback.
1 reply
-
DreamBooth is for subject-driven generation and NOT for generalization to new objects. The task is to generate a subject of interest in newer renditions. |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
When I trained playground2.5 on my own data set(Contains a series of images and their corresponding tags), I couldn't get good results. But when I use the base model of SD XL, it can be trained very well. What is the reason for this?


Beta Was this translation helpful? Give feedback.
All reactions