i have run this code, the cosine metric learning method in this code does not match that in the paper, and the performance is much worse than the author proposed in the paper, anyone has the same issue? and although they give the code of gaze loss, but seem like they didnt use it.
i have run this code, the cosine metric learning method in this code does not match that in the paper, and the performance is much worse than the author proposed in the paper, anyone has the same issue? and although they give the code of gaze loss, but seem like they didnt use it.