train

[yaningjm]

Why is the following error always reported when training the NTIRE2023 data set: Out of memory

[YuanGao-YG]

This error may be caused by insufficient GPU memory. Please try using a GPU with larger memory or reducing the size of the training image.

[yaningjm]

Thank you for your reply. The experiment using NVIDIA RTX A6000 50G video memory will also report an error of exceeding the memory. I guess there is a problem with the framed part in the picture, because the NTIRE2023 data set is too large and the test part cannot be loaded. Have you encountered such a problem during training? If so, can you tell me how to solve it?

@YuanGao-YG

[gy65896]

You are right. During testing, we use a scaled-down version for testing. If you want to test on the original size, you need A100 80G for testing.

[yaningjm]

Ok, I see. Thank you very much