Modifiable patch size in pretraining.

[MarkoHaralovic]

Thank you for sharing your work!

I tried changing the mask size inside configs/openmixup/pretrain/a2mim/imagenet100/<python_script_name>.py, but I continuously encountered an error. Specifically, I changed mask_patch_size to 8 and model_patch_size to 8 as well, but I received the following error:

RuntimeError: shape '[batch_size, 1, 2, 2]' is invalid for input of size batch_size*batch_size*2

Did you plan to implement the code so that the patch size is configurable?