How can I train PETRv2 in "current frame" only mode?

[Jeong-Bin]

Hello. Thank you for your research and sharing.

I want to train the PETRv2 segmentation model using only the "current frame" without the past frames.

In this case, should I just modify sweeps_num=1 to sweeps_num=0 in this code petrv2_BEVseg.py?

(+) I wonder how many past frames you use for PETRv2 segmentation.
Is it correct that sweeps_num=1 means "the model uses 1 past frame" as a default?
However, I found that you mentioned "2" from here.
Dose that "2" mean "1 past frame + 1 current frame"?