Extreme quality and variation model version but could be unstable.

Settings:

Steps - 29400

Dataset - 1024x1024

Images - 294

Network Module - LoRA

Base Model - flux1-dev-fp16

Trigger word - Jinx

Repeat - 1

Epoch - 100

Seed - none

Clip Skip - none

Text Encoder learning rate - 0.00001

Unet learning rate - 0.0001

LR Scheduler - constant

Optimizer - AdamW8bit

Network Dim (Rank) - 64

Network Alpha (Rank) - 64

Gradient Accumulation Steps - none

Shuffle caption - none

Keep n tokens - none

Noise offset - 0.03

Multires noise discount - 0.1

Multires noise iterations - 10

conv_dim - none

conv_alpha - none