How to implement standard deviation in deepstream

I have a model, trained using the PyTorch framework. This model uses mean and standard deviation for normalization before inference.
I know how to pass the mean file but How to pass standard deviation for each channel to the nvinfer?