This inversion of Jess from (the probably long defunct) KHA Models was done thusly:

Images: 16

Rate: 0.05:10, 0.02:20, 0.01:60, 0.005:100, 0.002:200, 0.001:400, 0.0005

Steps: 1220

Batch: 2

Wet hair in two of the dataset pics might make it a bit unruly here and there but they were the best shots of her face.