Stable Diffusion 3 Medium Fine-tuning Tutorial — Stability AI

https://stability.ai/news/stable-diffusion-3-medium-fine-tuning-tutorial

12 points · 8 comments · view on lemmy.world

8 Comments

istanbullu@lemmy.ml · 3 pts · 1y (7 replies)

Is a reason to use SD3 when Flux is much much better, with a much better license?

Even_Adder@lemmy.dbzer0.com · 1 pts · 1y (6 replies)

I don't think so. They're going to have to do a lot better than a tutorial to win people back. That said, the two Flux models being distilled making them close to impossible to fine-tune sucks too.

istanbullu@lemmy.ml · 1 pts · 1y (1 reply)

kohya now supports flux fine tuning. I have seen nice examples in civitai.

Even_Adder@lemmy.dbzer0.com · 0 pts · 1y

Those might just be LoRA merged models, not full fine-tuning. From what I heard, fine-tuning doesn't work because the models are distilled. You'd have to find a way to undistill them to train them.

clb92@feddit.dk · 1 pts · 1y (3 replies)

People have been training great Flux LoRAs for a while now, haven't they? Is a LoRA not a finetune, or have I misunderstood something?

Even_Adder@lemmy.dbzer0.com · 0 pts · 1y (2 replies)

Last I heard, LoRAs cause catastrophic forgetting in the model, and full fine-tuning doesn't really work.

clb92@feddit.dk · 2 pts · 1y

Oh well, in practice I'll just continue to enjoy this (possibly forgetful and not-fully-finetunable) model then, that still gives me amazing results 😊

erenkoylu@lemmy.ml · 0 pts · 1y

quite the opposite. Lora's are very effective against catastrophic forgetting, and full finetuning is very dangerous (but also much more powerful).