무료로 가입하고 즉시 10,000 크레딧을 받아보세요.
가입 당일 계정에 지급되는 신규 크리에이터용 시작 크레딧입니다.
사용자 LoRA
f0rg3
About this version Trained on Illustrious 2.0 stable
Optimizer: Prodigy
Scheduler: Cosine
Dataset size: 928 images
Steps: 13900 (?), In theory it should be less than 10,000, the civitai LoRA trainer is a bit broken..
Dim: 32 Alpha: 16
This is my attempt at training a LoRA on Forge Walfhammer's style, though the main focus here is actually on the anatomy.
Dataset size: 928 images.
If you want to get more realistic anatomy in some of your gens, try adding belly to your prompt.
Recommended Settings:
Weight: 0.8 – 1.0 (It doesn't seem to be overtrained, so a weight of 1.0 should work great for most setups).
Samplers: DPM-style samplers (like DPM++ 2M) tend to give the best style adherence, especially when it comes to rendering fine details like body hair.
