Best way to normalize the input?

not sure if normalize is the correct term here. The issue: The input for my neural network has different dimensions ranging from 1e-2 and 1e3.
What is the best way normalizing it ( for the forward call as well)

I looked into Batch Normalization, which expects it to be several data sets.

I just have one “dimension” which I want to bring into the same range of values.

You could calculate the mean and std of the entire dataset for each feature and normalize each input by subtracting the mean and dividing by std. torchvision.transforms.Normalize applies the same normalization using the mean and std values for each channel.