How the dim parameter of the Softmax() is reflected in CrossEntropyLoss()

[ptrblck]

nn.CrossEntropyLoss expects the class dimension in dim1 as explained here.