Patch extraction stride during evaluation

[TalRub104]

Hi,

In the paper, you mentioned that during training, you used a patch extraction stride of 16, and during fine-tuning, you used a stride of 8. From the GitHub, I understand that you are using a patch extraction stride of 16 for evaluation. Is it problematic to use a stride of 16 after fine-tuning with a stride of 8? The model was fine-tuned using a stride of 8, but now you're processing the data with a stride of 16.

Thanks