Not able to train the SegVit on the dataset with an input size 512x1024 

Thank you for uploading the code for your interesting project.

We are trying to retrain your SegVit code using our dataset. Is it possible to change the network so that input images with an aspect ratio of 1:2 are also supported? Thanks for your answer in advance. 