I want to know the parameter running_var in the batch normalization refers to the variance or the standard deviation? Because I want to use the model pretrained in Matconvnet to init the pytorch model.
From the name I would guess variance.
Yes, it’s clear. Thank you.