Batchnorm for N*C*L

my tensor shape is 8 X 5 X 512 (N X C X L) and i want to apply batchnorm on L how to do it??

You should try using torch.nn.BatchNorm1d(). Please find the relevant info on this pytorch page.

batchnorm1d normalize over C in N X C X L

You could permute the input tensor and swap C with L before applying the batchnorm and swap it back afterwards.