VIPL-Audio-Visual-Speech-Understanding/LipNet-PyTorch

CTC loss becomes Inf after a very first number of steps

trungtv opened this issue · 0 comments

Dear,
I am fine tuning your model for our domain dataset. However, after a few steps, net(y) becomes nan and so CTC loss is Inf. Could you please give me any insight on what to do?
Thank you very much,