Help fuel my passion! Even $1 makes a difference. ☕ https://ko-fi.com/citronlegacy
Please checkout all my Pokemon Loras: https://civarchive.com/user/CitronLegacy/models?tag=pokemon
This Lora can make Togepi and also a person holding Togepi. Misty is trained into the Lora can be prompted with tgpiMsty other girls can be holding Togepi by prompting holding tgpi
Description
Trained on: 73 images
Training Model: Animefull-final-pruned-fp16
flip_aug: False
Num of Repeats: 10
Unit is Epochs or Steps: Epochs
Number of Epochs or Steps: 10
Training Batch Size: 2
Total Steps: 3650
Resolution: 512
Network Dim: Dim 32 and Alpha 16
Lora Creation Process took: 00:40:55
FAQ
Comments (2)
Hey is this for the bounty? If you don’t mind me asking do you have any tips on getting the best results for a model like this? I made one too but I don’t think it’s as good as this. I had about 24 training images and I ran 80 dataset repeats at 2500 steps I’m still pretty new so those are really the only settings I touch
Hey! I like your Maple Story Slime avatar!
Yes I made it for the bounty. Making accurate Pokedex Loras is of my specialties.
I'm happy to provide tips/tricks!
Yours didn't look bad.
Tips:
1. Check out the "About this version" menu in the right sidebar. It lists my training settings. (I always list them for SD1.5 and occasionally for XL)
2. You can see all the settings I used with CivitAI trainer and some extra notes I made about how many repeats to use. I had to make this PonyXL Lora twice to get it correct.
3. Generally when training a PonyXL model on CivitAI I take the defaults and change only 4 settings:
-- Number of Repeats
-- Batch size
-- Shuffle Caption: true
-- KeepTokens: 1