ELLA
ELLA copied to clipboard
Questions regarding ELLA training
Thank you for the amazing work and for sharing this project with the community! I have been trying to train ELLA for SDXL, and I have a few questions regarding the training setup and methodology, and I would greatly appreciate your clarification:
- Number of Epochs: How many epochs did you train the model for? and what was the batch size used for training?
- EMA - Did you utilize EMA for model weight updates?
- AI artifacts - I am training on 512 resolution, and I am getting lots of distorted objects and AI artifacts in my images. How did you solve for this?
- How did the model perform when conditioned on short prompts? I look forward to hearing from you!