Could weight_g in weight_norm become negative?

1

weight_g is from torch.nn.utils.weight_norm, Pytorch 1.1.0
During training, could weight_g become zero or negative?
I can not find any document