1. expand(size)
Example:
x = torch.rand(1,2,3)
print(x)
print("x.shape:", x.shape)
y = x.expand(2, 2,3)
print(y)
print("y.shape:", y.shape)
Output:
tensor([[[0.8020, 0.2300, 0.8632],
[0.9463, 0.0172, 0.9756]]])
x.shape: torch.Size([1, 2, 3])
tensor([[[0.8020, 0.2300, 0.8632],
[0.9463, 0.0172, 0.9756]],
[[0.8020, 0.2300, 0.8632],
[0.9463, 0.0172, 0.9756]]])
y.shape: torch.Size([2, 2, 3])
Note
expand only on the number of dimensions in tensor is operated to be extended to a dimension of, for example, the example above only the first extended dimension, if the change x.expand(1, 4, 3)
being given, expand_as line presentation face and expand almost the same, expand_as and out parameter is a tensor
2. tensor.expand_as(tensor1)
Example:
x = torch.rand(1,2,3)
z = torch.rand(3, 2,3)
print(x)
print("x.shape:", x.shape)
y = x.expand_as(z)
print(y)
print("y.shape:", y.shape)
Output:
tensor([[[0.5696, 0.3571, 0.2565],
[0.9646, 0.5754, 0.7819]]])
x.shape: torch.Size([1, 2, 3])
tensor([[[0.5696, 0.3571, 0.2565],
[0.9646, 0.5754, 0.7819]],
[[0.5696, 0.3571, 0.2565],
[0.9646, 0.5754, 0.7819]],
[[0.5696, 0.3571, 0.2565],
[0.9646, 0.5754, 0.7819]]])
y.shape: torch.Size([3, 2, 3])