NegLogProbLoss.scale# NegLogProbLoss.scale: bool# If True, divide losses by the training sample size.
