pytorch repeat()函数的功能演示，非常详细（代码验证）

PyTorch中的repeat()函数可以对张量进行重复扩充。

repeat（）中的参数个数需 >=  tensor 维度

举例证明：

>>> abb = torch.rand([2,1,2,3])
>>> print("abb: ",abb)
abb:  tensor([[[[0.1103, 0.8384, 0.8689],
          [0.2282, 0.9317, 0.4930]]],


        [[[0.8018, 0.8036, 0.1346],
          [0.9017, 0.4192, 0.6952]]]])
>>> print("abb.shape: ",abb.shape)
abb.shape:  torch.Size([2, 1, 2, 3])
>>> bb = abb.repeat(1,2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RuntimeError: Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor
>>>

看看 错误原因：RuntimeError: Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor.

当参数只有两个时：（行的倍数，列的倍数）， 此时通道数倍数为1，表示不重复扩充。

当参数有三个时：（通道数的倍数，行的倍数，列的倍数）。

直接将repeat参数对应乘上原tensor的shape就得到最终的shape，如：
shape(2,3) * repeat(1,2) = shape(2,6)  //channel = 1

shape(2,3) * repeat(2,2,1) = shape(2,4,3)  // shape(2,3) 可看作shape(1,2.3)
 

>>> bb = abb.repeat(2,2,2,3)
>>> bb
tensor([[[[0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930],
          [0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930]],

         [[0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930],
          [0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930]]],


        [[[0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952],
          [0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952]],

         [[0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952],
          [0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952]]],


        [[[0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930],
          [0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930]],

         [[0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930],
          [0.1103, 0.8384, 0.8689, 0.1103, 0.8384, 0.8689, 0.1103, 0.8384,
           0.8689],
          [0.2282, 0.9317, 0.4930, 0.2282, 0.9317, 0.4930, 0.2282, 0.9317,
           0.4930]]],


        [[[0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952],
          [0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952]],

         [[0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952],
          [0.8018, 0.8036, 0.1346, 0.8018, 0.8036, 0.1346, 0.8018, 0.8036,
           0.1346],
          [0.9017, 0.4192, 0.6952, 0.9017, 0.4192, 0.6952, 0.9017, 0.4192,
           0.6952]]]])
>>> print("bb.shape: ",bb.shape)
bb.shape:  torch.Size([4, 2, 4, 9])
>>>