Bias with batchnorm

Shouldn’t bias=false when we use batchnorm with linear-layer as in the case of conv-layer?