Preprocessing problem

[LLimerence]

Hello, I'd like to ask what kind of window width is used for your pre training model?

Is it obtained in the image or the three different window widths in your paper.