This LoRA model is in testing stage and has some severe problems at this time. Do not expect it to be competent enough generating at best quality.
My first LoRA model, training with 108 illustrations and stopped at 2000 steps.
I am preparing for the next training today, so it should be update soon.
Description
param(
[string]$Optimizer = "customized_optimizers.muon_lora.AdaMuonLoRA",
[string]$LR = "5e-3",
[int] $MaxSteps = 2000,
[int] $Dim = 64,
[int] $Alpha = 64
)