import torch t = torch.tensor([ 1, 2, 3, 4 ]) print(t) # # tensor([1, 2, 3, 4]) u = t.expand(3, -1) # # expand() creates a view, therefore, changes to # t are reflected in u. # t[1 ] = 9 # # Because u is a view, changes to u are reflected # in t as well (and by extension in all rows of u): # u[2,3] = 8 print(u) # # tensor([[1, 9, 3, 8], # [1, 9, 3, 8], # [1, 9, 3, 8]]) print(t) # # tensor([1, 9, 3, 8])