👼 Her angelic voice rings a familiar tune.
Works at weight 1. Made with my Lora guide.
Description
Trained with 120 images of Nazuna. Download the dataset if you're interested.
FAQ
Comments (8)
Yo, sorry for bothering you again. Your answer on the Lora guide page was very helpful. I'm just curious what tag threshold did you use for tagging this dataset images? Cuz my Lora seems too strict in what it can do.
Also, forgot to ask. If i train for example with 10 repeats and 20 epochs and with 20 repeats and 10 epochs, will results vary? If yes, which is better? More epochs or more repeats?
P.S. What were your repeats and epochs for this Lora?
I used 0.5 and even removed some redundant tags.
More epochs is probably slightly better than more repeats, because it ensures the dataset is walked through in its entirety before using the same image. It also makes it noticeably harder to overcook.
You can see my settings by downloading the dataset. I used 120 images, 3 repeats and 10 epochs (picked the 7th)
Sorry to bother you! Could you please make one for Zentreya to separate both of her designs? I feel that every existing dataset for Zen is heavily polluted by both forms- giving her robot form horns and the likes even when you negative prompt them and so on.
I'd be extremely appreciative! Thank ya kindly sir!
I'm not a fan of Zentreya, but I will heavily consider it, thank you for your kind comment and thank you for the suggestion!
OOOOOOOYEESSSSSSSSSSSSS you did it !
I'm also trying Mikeneko but having a hard time
@Hollowstrawberry rushia, nazuna, mikenneko, yep i seeing a pattern here :p
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.