search

chengyangfu / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
169 stars 47 forks source link

loss abnormal #25

Open melon0504 opened 6 years ago

melon0504 commented 6 years ago

why the loss is nan while the loss of the same data training with SSD and RefineDet is normal

lhx150926 commented 6 years ago

I faced the problem too and the problem is occurred in the start of training

melon0504 commented 6 years ago

which model did you use, CPU or GPU?