LoRA

[JohnTigue]

is a second-gen fine-tuning method. One of its goals is to have more sharable (that is, smaller) models by training only the delta, not the full model. This also means the fine-tuning delta can be applied with a slider from 0%--100%. Although LoRA is early in development, this looks very promising.

• run_inference.ipynb
• Version 0.1.0 of LoRA released! (alternative to Dreambooth, 3mb sharable files)

[JohnTigue]

LoRA is definitely going to make waves:

• LORA for Stable Diffusion - A1111 Dreambooth Extension
• 8 images
• How To Do Stable Diffusion LORA Training By Using Web UI On Different Models

[JohnTigue]

How to extract small LoRA file from custom dreambooth models . Reduce your model sizes!:

Disclaimer : I tried on one custom dreambooth model only and it worked like charm. If more "Style" models and DB models can be extracted, it would be of tremendous value to reduce their filesizes.

[JohnTigue]

Olivio: LORA: Install Guide and Super-High Quality Training - with Community Models!!!

[JohnTigue]

MERGE A FACE & STYLE With LORA EXTRACTION In Stable Diffusion! NO TRAINING!

[JohnTigue]

LORA with ControlNET - Get the BEST results - Complete Guide // Jenna Ortega