The italian actress Monica Bellucci,
Lora trained on 768*768 resolution, but can be used on 512*512 models aswell.
Description
FAQ
Comments (4)
Congratulations!
Could you give instructions on how you trained the Lora? with what method, fine-tune, dreambooth, etc..
How many images or with what tutorial? Thank you so much!
Hi! im glad you like it 🙂
i trained the LoRa using Kohya_ss GUI,
THIS is the tutorial i followed to learn how install it and use it, and THESE are the values that i have parameterized.
To train this one i used 24 images of 768x768 res at 100 steps each one (which is 2400 steps in total not 1500 as i posted ). The training took around 30 minutes with a RTX 3060 12Gb vram. If i use 10 images of 512x512 like the tutorial that takes less than 10 minutes.
this is the first time i train with that amount of images and resolution, I wanted to test the time/quality.
i think you can get similar results with less images.
I hope you have cleared some doubts 😀
and sorry for my english 😝
@crabeon No te preocupes por el inglés. Estoy muy interesado en este tema y he estado haciendo pruebas. Utilizo googleColab por lo que no se peude entrenar a 768. Hice un tutorial como ese, pero no obtuve tan buenos resultados.
Podemos estar en contacto de alguna manera? Quizás necesito alguna ayuda más :)