jerryli comments

Results 6 comments of


                                            jerryli

error :overflow encountered in exp and multiply

got the same problem

RuntimeError: cannot reshape tensor of 0 elements into shape [0, -1] because the unspecified dimension size -1 can be any value and is ambiguous

> > > 在fire.Fire（）括号里加train试试加train 之后 Adam 可以，SGD 也报同样的错

RuntimeError: cannot reshape tensor of 0 elements into shape [0, -1] because the unspecified dimension size -1 can be any value and is ambiguous

> > > 在fire.Fire（）括号里加train试试还有，这个问题为啥可以加train试一下

RuntimeError: cannot reshape tensor of 0 elements into shape [0, -1] because the unspecified dimension size -1 can be any value and is ambiguous

> > > 您好，加上adam效果是不是没有sgd好啊。好像是，还没调好我

RuntimeError: cannot reshape tensor of 0 elements into shape [0, -1] because the unspecified dimension size -1 can be any value and is ambiguous

> > > > > 您好，加上adam效果是不是没有sgd好啊。 > > > > > > 好像是，还没调好我 > > 麻烦弄好能公布一个mAP吗？谢谢了。我debug找出的原因是在loc2box的函数中，在计算dw，dh时候exp溢出，进而导致RPN生成的128候选框的坐标变成nan。然后网络将这些候选框给裁掉之后导致不够128个。最终batch变成了0.报的错。然后，loss变成nan，梯度爆炸。好的呢

自动压缩评估时报错, 最后map很低

我也报同样的问题