Joycaption now has both multi GPU support and batch size support > https://www.patreon.com/posts/110613301
FLUX LoRA training configurations fully updated and now works as low as 8GB GPUs — yes you can train on 8 GB GPU a 12 billion parameter model — very good speed and quality > https://www.patreon.com/posts/110293257
Check out the images to see all details
