LyCORIS

Why we included this project

Most LoRA fine-tuning for Stable Diffusion eventually crosses paths with LyCORIS. The project bundles a range of parameter-efficient adaptation methods, not just plain LoRA: LoCon, LoHa, LoKr, DyLoRA, and (IA)^3, and it also supports native fine-tuning (DreamBooth). Its README includes a comparison table rating methods on fidelity, diversity, and size, which is useful when you need to choose one for a specific concept rather than guess. The guidelines explain the reasoning behind each variant, so when a LoRA starts overfitting or collapsing you have somewhere to look for an explanation. For teams wiring up their own fine-tuning pipelines on SD1.x, SD2.x, or SDXL, it is a maintained, paper-backed reference implementation instead of a throwaway script.