Thanks for your nice work!
But when I retrained your searched architecture with default DARTS settings, I got the following results (the same architecture with 8 runs):
97.379997
97.309998
97.399998
97.489998
97.389997
97.419998
97.269997
97.229997
In your paper, the reslut is 97.68. I suspect my training skills may be wrong, but I wonder how to train the model so high? So could you please provide your log file or the random seed?
Great thanks and best wishes!
Thanks for your nice work! But when I retrained your searched architecture with default DARTS settings, I got the following results (the same architecture with 8 runs): 97.379997 97.309998 97.399998 97.489998 97.389997 97.419998 97.269997 97.229997 In your paper, the reslut is 97.68. I suspect my training skills may be wrong, but I wonder how to train the model so high? So could you please provide your log file or the random seed? Great thanks and best wishes!