4uiiurz1 / pytorch-adacos

PyTorch implementation of AdaCos
MIT License
211 stars 32 forks source link

softmax activation on very last layer #14

Open gihyunkim opened 4 years ago

gihyunkim commented 4 years ago

is it not supposed to use softmax function on last layer? I am confused that someone use softmax on last layer and others don't. Does it gives different result whether I use softmax or not? or it just for convenience to evaluate. plz let me know

smile-cpu commented 1 year ago

使用交叉熵损失函数,交叉熵损失函数自带softmax()