RDBT | MOSS
RIP anima (?)
This is a finetuned LoRA from a personal all-in-one dataset, mainly for digital art and drawings.
LoRA is trained against MOSS base (not sft). Cover images are also from MOSS base. Also works on SFT, but you might need lower strength.
Still testing. I might need to change the way to caption, or just train dedicated LoRAs. Styles seem to be messed up all together.
For now, I recommend to use it as a "regulator/stabilizer/detailer/whatever" LoRA with strength (0.2~0.5). It can alleviate the overfitting from other style LoRAs.
Or higher strength if you like the style.
FYI, about training, if someone interested:
ai toolkit supports comfy int8 ckpt. 16gb vram GPU should be fine. 12gb is questionable. It allocated ~12.5gb vram on my end.
3x slower than Anima, acceptable.
I use bs 8 (accumulated) and lr 0.00005.
v0.1: This is a test run.
Sharing merges using this model is not allowed. This "restriction" won't affect anyone. It's only aimed at those who steal others' models to sell.
Description
Comments (2)
style loras seem to perform better in combination with Lodestone's 8step dmd lora (just use a cfg of 2 or higher though and negative prompts)
This is promising, boss. Unlike the other general purpose, larger models, this is one seems to have actually been trained on the dark arts of the booru bible. Besides your usual magic on stabilizing, I'm hoping you can get it to match Anima or IL when it comes to style fidelity. It already seems to have some knowledge there, so maybe it just needs a little nudge.