search

hollowstrawberry / kohya-colab

Accessible Google Colab notebooks for Stable Diffusion Lora training, based on the work of kohya-ss and Linaqruf
GNU General Public License v3.0
561 stars 80 forks source link

Add v parameterization option #84

Open Fu0x opened 5 months ago

Fu0x commented 5 months ago

Hi, is there a way to enable V-parameterization for training Loras on a terminal-snr-v-prediction model? Would really appreciate it!

(btw. i am talking about a non sd-v2 model)