search

CasiaFan / tensorflow_retinanet

RetinaNet with Focal Loss implemented by Tensorflow
121 stars 45 forks source link

loss value #1

Open vcca opened 6 years ago

vcca commented 6 years ago

I didn't use the one-hot encoding, but my loss got be NAN.

kehaozhe commented 6 years ago

I got the same loss 'NaN'. At the beginning, I didn't use the one-hot encoding labels, got NaN, After I changed my label to onehot encoding, got NaN too.

vcca commented 6 years ago

try to modify the learning rate to a smaller value

kehaozhe commented 6 years ago

I got Nan loss in first iteration~