A small running error

[simple-boy]

I applied cornernet Lite to my dataset, which has 28 classes and the size of each image is 5472 * 3648. After training, I use demo.py for a test. However, only when the size is modified to a smaller size when the test picture is read, can it operate correctly. When the size of the picture is not modified, there will be an error, "Runtimeerror: cudnn error: cudnn_status_international_error".

How can I modify it to run completely without changing the size of the picture.

Thanks for your help.