how to avoid getting nan in loss function?

Question

I\'m getting NAN in loss function and I gave checked that there is no NA in dataset, please find the below code for model

keras.backend.clear_session()
input1 = I