Resolution of rendered video not matching the original video if size is not multiple than 16

[heimrabe]

Hi thank you again for your work. I have noticed that when running the training on a custom monocular video of resolution 1920x1080 the rendering I get afterwards has size 1920x1072. Looking through the code I think it is caused by the fact that 1080 is not a multiple of the default tile_size of the gaussian renderer, which is 16. Do you know how this impacts the reconstruction? Is it cut from the top or the bottom? Is there a way to pass this parameter through so that the rendered video is also 1920x1080?Also, does this have an impact on the training?