Exclude weights don't improve performance of neuralnet and produces NaN

Question

How do I exclude weights using neuralnet in R and make predictions that don\'t result in NaNs? Predictions are produced correctly when I remove the exclude