Returns a contiguous tensor containing the same data as self
tensor. If self
tensor is contiguous, this function returns the self
tensor.
普通的张良在内存中都是连续存放的,然而有些张量经过一些处理后(如permute)后就不是连续的了,此时这些张量无法用基于连续内存空间的内置函数,如view
a=torch.randn(2,3,4)
a=a.permute(2,1,0)
print(a.shape)
>>>torch.Size([4, 3, 2])
a.view(2,1,-1)
>>>RuntimeError
Traceback (most recent call last)
<ipython-input-41-e061ce4ab316> in <module>()----> 1 a.view(2,1,-1)
a=a.contiguous()
a.view(2,1,-1)
>>>tensor([[[-0.5943, -0.4082, -0.0357, -1.5612, -0.4275, -1.8975, -0.4403,
0.1456, -0.8696, 0.8224, -1.1907, 0.8248]],
[[ 0.8431, 0.0476, -0.2804, -0.7271, -0.4674, 1.9256, 0.1905,
0.7308, 1.5797, -3.1329, 1.1580, -0.2556]]])
而contiguous()类似于clone,返回一个有连续内存的self张量的copy