Description
FAQ
Comments (6)
Can do 3dkx 1.1?
Can you tell us how to extract LORAs in different quality: 256, 320 ? This significantly reduce the filesize of the models. Is there a script publicly available, maybe on github ?
Supermerger extension for the automatic1111 web gui and Kohya SS currently allow the extraction of LoRAs
@daemonrat Thank you !
@ritcher1 download the noise offset here to mix with https://civitai.com/models/10391/noise-offset-for-true-darkness-in-sd
Provides higher fidelity
@daemonrat Found 1 good explanation page:
"You don't really need the whole giant kohya distribution for that(it pulls with it several GBs of downloads!), just download two files from kohya_ss\networks: lora.py and extract_lora_from_models.py
and make a .bat file to execute them:
python.exe extract_lora_from_models.py --save_precision fp16 --save_to "result_name.safetensors" --model_org "model_base.ckpt" --model_tuned "model_tuned.safetensors" --dim 32"
And another interesting script:
https://github.com/sashaok123/LoRA-EXTRACTOR
And an interesting comparison discussion "How good is LORA extraction" ?
https://www.reddit.com/r/StableDiffusion/comments/10zh1fw/lora_extractions_vs_db_model_can_we_truly_replace/