Question regarding `calculate_means`

[letme-hj]

Hi, thank you for your work and sharing your code!

While digging in your code, I found that CUB dataset and the others have different implementations of calculate_means. https://github.com/zhukaii/SPPR/blob/0df749d000e50a64ae13c606072a902f19ecb251/utils/model/my_model.py#L156-L169

As far as I understood, for classification, cub brings its base class prototypes from the parameters that were updated during the base session, and the others are directly using the output of the attention module (IL_attn).

I wonder why there is such difference across the datasets in calculating base class prototypes . Or if I misunderstood something, please let me know!

Thank you