Strange result when training

[uracilForever]

I modified the code and tried to train on my dataset,but the result is strange that the loss function was increasing rapidly and then became NAN.
I had checked out the dataset and didn't find any mistake...what should I do ...