Pruning with weight_norm

Hi, when using weight_norm we replace weight tensor with weight_g and weight_v tensors. While pruning which tensor should be masked? Should mask just weight tensor or weight_g and weight_v?