Question about fern_10v in llff dataset

[IncludeMathH]

Thanks to the author for this awesome work. I'm having some trouble reproducing the training process, namely that fern_10v contains images that are too large in resolution. Other than this subset, which is typically 756 X 1008, this subset is 3024 X 4032 in size, so much so that OOM occurs when rendering the test set images.
I'd like to ask the author how to deal with this problem. Can you reduce the images to 756 X 1008 and rerun the colmap?