view()和reshape()比较

PyTorch Tensor自带的方法view()和reshape()：

• view(): Returns a new tensor with the same data as the self tensor but of a different shape.
• reshape(): Returns a tensor with the same data and number of elements as self but with the specified shape.

使用下来，都是作用都是更改Tensor的Shape，而且性能没啥差别 view vs reshape

但是，view要求tensor元素连续，若不连续，则推荐用reshape

Traceback (most recent call last):
File "d:/pytorch/tensor_reshape.py", line 9, in <module>
print(y.view(9))
RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension
spans across two contiguous subspaces). Use .reshape(...) instead.

import torch

x_3x3 = torch.arange(9).reshape(3,3)
print(x_3x3)

# Transpose
y = x_3x3.t()
print(y)
# view size is not compatible with input tensor's size and stride (at least one dimension
# spans across two contiguous subspaces). Use .reshape(...) instead
# print(y.view(9)) 
print(y.reshape(9))

总结：优先选用reshape， 而不是view