giagrad.Tensor.permute

Tensor.permute(axes=None)

Returns a view of the original tensor with its axes permuted.

permute uses numpy.transpose, the following documentation is adapted.

For a 1-D tensor, this returns an unchanged view of the original tensor, as a transposed vector is simply the same vector. To convert a 1-D tensor into a 2-D tensor vector, an additional dimension must be added, e.g., tensor.unsqueeze(axis=0) achieves this, as does Tensor[:, None].

For a 2-D tensor, this is the standard matrix transpose. For an n-D tensor, if axes are given, their order indicates how the axes are permuted (see Examples). If axes are not provided, then tensor.permute().shape == tensor.shape[::-1].

See also

giagrad.Tensor.swapaxes()

For swaping only two axes.

numpy.transpose()

Parameters:

axes (tuple or list of ints, optional) – If specified, it must be a tuple or list which contains a permutation of [0,1,…,N-1] where N is the number of axes of the original tensor. The i’th axis of the returned tensor will correspond to the axis numbered axes[i] of the input. If not specified, defaults to range(tensor.ndim)[::-1], which reverses the order of the axes.

Examples

>>> t = Tensor.empty(1, 2, 3, 2, dtype=int).uniform(-5, 5)
>>> t
tensor: [[[[ 1  0]
           [-3  4]
           [ 3  3]]
...
          [[ 3 -4]
           [-3  1]
           [-2  3]]]]

Note that axes has the same lenght as giagrad.Tensor.ndim.

>>> t.permute(axes=(1, 2, 3, 0))
tensor: [[[[ 1]
           [ 0]]
...
          [[-3]
           [ 4]]
...
          [[ 3]
           [ 3]]]
...
...
         [[[ 3]
           [-4]]
...
          [[-3]
           [ 1]]
...
          [[-2]
           [ 3]]]] fn: Permute(axes = (1, 2, 3, 0))
>>> t.permute(axes=(1, 2, 3, 0)).shape
(2, 3, 2, 1)