patch-wise model could only reach 68%

[minqz2009]

I've tried to run the patch-wise model for multiple times, but could only reach a maximum accuracy of around 70% and average accuracy around 60%, here's all the hyperparameters I've used:


The result has a huge gap towards the released result in the paper, did I do something wrong? or the released result is indeed a very rare result?

[giao-cloude]

I also have the same problem. Have you solved it