impression2805 / il2a Goto Github PK

Hi, Thank you very much for the code base!
I am reproducing this article using the cifar100 dataset without making any changes to the code on this github, why the results are a bit different from your paper, do I need to adjust some parameters?
It would be great if you could solve my problem, thanks!

Thank you very much for the code base!
I am having troubles reproducing the results in the paper. In section 3.1.1 of the paper, you introduced a spectral decomposition method for analyzing forgetting. But I got confused about the format of the deep features, I want to know whether it is a feature directly obtained in a feature extractor (like [160, 4, 4]) or a feature vector after flattening (like [2560, ]).
It would be great if you could provide some more details on the above issues.
Thank you very much!