caffe 运行出错的问题记录

问题1：

1、 Checked failed shape[i]>=0

http://stackoverflow.com/questions/37683320/caffe-imagenet-checked-failed-shapei-0

原因：1 卷积核大小等设置不对

            2 数据集制作不对

2、Common causes of nans during training

训练时候出现无穷 

http://stackoverflow.com/questions/33962226/common-causes-of-nans-during-training

原因 ：

1 

Gradient blow up

Reason: large gradients throw the learning process off-track.

2

Bad learning rate policy and params

3

Faulty Loss function

4

Faulty input

5

stride larger than kernel size in "Pooling" layer

6

Instabilities in "BatchNorm"