LoRAs are trained on 32 different images, mostly on fanarts and images from Swooning Over Stans
Preview generated without using ADetailer & Hires. fix
Recommended LoRA Strength (Weight): 0.7-0.8
v1.0 NoobAI-XL EPS 1.0/Illustrious-XL v0.1 - v1.0 the same as NEW Edited (Pony)
v1.0 NEW Edited - In this version, the images were edited to improve the creation of glasses, the captions were also slightly edited
v1.0 NEW - A small update for v1.0 - Captions have been changed and improved, Trigger Word has also been changed, Some images have also been improved. Images with a style similar to Gravity Falls have been marked with the toon\(style\) token, so use this token to make the style look like GF and do not use score in this case! (31.08.2024)
v1.0 (MT) - (MT) More Training. This model differs only in the number of steps, 3200 steps instead of 1600 steps (14.03.2024)
v1.0 - This model is trained with 1600 steps, in my opinion MT works a little better than this version (22.01.2024)
Description
Repeats: 10
Epoch: 5+10
Image Count: 32
Network Rank (Dimension): 16
Network Alpha: 8
Batch size: 1
Clip skip: 1
Training steps: 1600 - 5e, 3200 - 10e
Resolution: 640x960
Train Resolution: 1024,1024
Model: ponyDiffusionV6XL_v6
LR Scheduler: cosine_with_restarts
Optimizer: AdamW8bit
Learning rate: 0,0005
Text Encoder learning rate: 0,00005
Unet learning rate: 0,0005
Captioning: WD14 Captioning (SmilingWolf/wd-v1-4-moat-tagger-v2) + Edit
Recommended LoRA Strength (Weight): 0.7-0.8
Trigger Words: stanford pines
Merging of the 5th and 10th epochs