My first LoRA, so I didn't know what I was doing. Turn out Ok though.

I didn't know what alpha was so I used Dim: 2 Alpha:32, somehow that did not cause the model to overfit?!

NWM.Watercolor.Style.2024-11-02.03.48.52.png

2024-11-02 03:48:52

FLUX.1 - dev-fp8

Trigger: nwmwc1

Repeat: 20 Epoch: 10

Unet learning rate: 0.0001

LR Scheduler: constant

Optimizer: AdamW8bit

Network Dim: 2 Network Alpha: 32

Prompt: 1girl cowgirl in white short shorts, field with blue sky

Sampler: euler

Epoch Loss

1 0.338

2 0.336

3 0.332

4 0.328

5 0.321

6 0.309

7 0.310

8 0.298

9 0.297

10 0.281