Batchnorm has other parameters - in addition to `weight` and `bias` - that are not counted in torchsummary, such as `running_mean` and `running_var`: https://pytorch.org/docs/stable/_modules/torch/nn/modules/batchnorm.html