Reproducing the performance of the face recognition model reported in the paper.

[too-sheep]

Hi, Congrats to the authors! Great work, the performance of the synthetic data is almost on the verge of exceeding the real data. I'm having some problems reproducing the paper. The performance in Table 2 can't be reproduced, could you provide the parameters used to train AdaFace on CASIA-WebFace and the synthetic data?
Thank you for your efforts!

[Oguzhanercan]

Same problem, also face rec results on generated faces are really bad. Rank 1 is around %8, InstantID performs better, like above %80