Train SRGAN with high-resolution images of different dimension

[MattiaMartinelliGeo]

Good morning, thank you very much for the grate work. I am running the code of the SRGAN. I was able to run the code using high resolution images all of same size (e.g. 240x240 pixels), changing hr_shape and vgg shape values into the code. I was wondering if it is possible to train the mode using high resolution images of different sizes, for example some with dimension of 240x240, ohers of 100x100 etc

thank you very much