strange loss and presion

[3-leaves-grass]

@ZhaoJ9014 I used the default parameters in your config and trained model in the dataset（Asian-Celeb 93,979 ids with 2,830,146 aligned images)

，however， the loss was still high and Precision was 0 after epoch 72，what should I do ？ is this normal？or is it the matter of parameters in config？

[ShaoLingXie]

hi, i also meet this problem, do u solve it?

[3-leaves-grass]

yeah， making the value of s smaller can solve it

[pppnkn]

Hello, I have the same problem as well. Could you please tell me what is the value of s ? @3-leaves-grass thank you.

[3-leaves-grass]

the parameter s is the norm of input feature(default 64) , can find it in metrics.py @pppnkn

[XinWangg]

Hello, I have the same problem as well. Could you please tell me how to set the value s? what is the value of s you choose?