Pretty forgiving with weights, I think .9 works well.
Description
Same training as v2, only the lora size is doubled. Seems to produce less duds.
FAQ
Comments (1)
V3 bounced back at me with an error when I loaded SD. "not a safetensor"
AssertionError: I:\stable-diffusion-webui\models\Lora\Elina Karimova\ElinaKarimova_v3.safetensors is not a safetensors file