Fine-tuned level4 with 100s of images. Trained on 768x768. Use 'sks' in prompt. Check out more at patreon.com/Virtualbeautiesai and https://www.instagram.com/ai_beauty_go/
Description
FAQ
Comments (3)
Hi! Is there a particular reason you chose to use a checkpoint vs a lora ? Do you achieve higher quality by using a 768 trained checkpoint vs lora? Or is there some other thing that I'm missing?
I’ve only played with loras a little bit - extracting them from existing trained models. If I do that and then use with the same base model I get great results but if use with any other model they stink. So I don’t see a huge point (since HD space is not an issue for me. Mostly I dislike how much Lora’s usually distort the basic style of an imagr
@Walters1984 Thanks for the response. I only recently delved into fine-tuning models and doing lora extraction from them, and I think i'm seeing what you are regarding better quality output. Thanks.