Reason for downsampling in inference.py

[daeunni]

Hi, thanks for your nice work.
Btw, what's the reason for downsampling the original width/height?
Is it for sending to latent space? (i.e. VAE scale factor?)
Then how can I change it in pixel space? (w/ user-demanded height, width)

    ## latent noise shape
    h, w = args.height // 8, args.width // 8

https://github.com/AILab-CVC/VideoCrafter/blob/11bcd76fc62fb98b9715b994afe45b4fa081120c/scripts/evaluation/inference.py#L57C5-L57C45