Image focus, resolution and settings about Style training

[SheZ0322]

hello, I want to train a anime style model on 250 images, there are some things I want to ask:

1. Should I crop the full body image (rectangle) into half body or just a portrait. I used to resize it and make full body in 512 size, but it seems always get bad faces when I generate full body picture. Is it because of low resolution of face? Can I slove it by cropping them all into portrait? Should I focus on face, body or both? Will it influence the style I trained?
2. I once trained full body on 1024 size, and it can get good faces. So is higher resolution training always better than lower?
3. Is sd 2.1 better than 1.5?
4. To train this style, how do I change the settings? (like Ir, steps, concept img or sth)
5. When should I use Style_Training option?
   Thanks a lot!