# First look at the size of the incoming tensor, the incoming tensor's dimension is not enough, then you need to add dimensionality
# Here the size of the labels is tensor([n]), pass in the need to add dimensionality
# If the above problem occurs, just add to.(torch.int64) at the end to solve it
# n is the kind to be divided
labels = torch.nn.functional.one_hot(labels.unsqueeze(0).to(torch.int64), n)

Pro test valid!!!