Why My CTC Loss is always Nan or Inf value?

[hemanthyernagula]

While training the model at every epoch the loss is always a nan or inf value, I tried to debug but could not sort out. If anyone face this kind of issue please help me out.

[zuoyuwei]

encounter the same question, ask for help

[RobertXiaobinW]

Have you already solved the problem?

[gittjob]

I'm in the same situation. Have you solved it?

[hemanthyernagula]

No I did not

[GraceKafuu]

Same problem

[Yuxiang1995]

if you use your own dataset:

1. make sure all the labels are not empty
2. make sure the length of the predicted sequence > 2 * length(label) + 1