When use torch.op and when to use tensor.op for tensor?

[fangyh]

For torch.op:

torch.unsqueeze()
torch.reshape()
...

For tensor.op:

tensor.view()
tensor.repeat()
...

Is there any rule for these?