Issues about the performance of vanilla EVA-02-S

[jwfanDL]

Thank you for your amazing work.

I notice that your model architecture is slightly different from ViT-S or DeIT-S. However, the paper doesn't provide the performance of EVA-02-S trained by prevalent training strategies (i.e. AdamW, 300 epochs, RandAug + RandErasing ).

I would be appreciate if you can provide these results.