Super Resolution algorithms for very large images

Hello guys
I’ve been wondering how is it possible to apply a super resolution model to very large images (2000x2000)
I know and use pytorch distributed parallel training, but was wondering if there was something similar to it for the test
The test code I’m trying to run is the following’s :

If the image I try to upscale surpasses a resolution (600x600) it is not possible to test on it because of out of memory issue.
In terms of graphic prowess, I work on a server with multiple nvidia Tesla gpus on multiple nodes so I can use as many gpus as possible.
Thank you

You can split into several small images for processing and then stitch them together.