To which resolution do you usually crop your images when training LoRA models?

Poll options
What's this?