pytorch/torch中的view

类似于numpy中的reshape，且不改变原始数据。重点是，它是从最后一个维度进行reshape中的

import torch
a = torch.randint(0, 10, size=(3, 4))
print(a)
print(a.view(4, 3))
print(a.view(1, 12))

分别得到

tensor([[7, 1, 7, 2],
        [9, 5, 7, 1],
        [6, 0, 9, 9]])
        
tensor([[7, 1, 7],
        [2, 9, 5],
        [7, 1, 6],
        [0, 9, 9]])
        
tensor([[7, 1, 7, 2, 9, 5, 7, 1, 6, 0, 9, 9]])