chainer.functions.transpose

chainer.functions.transpose(x, axes=None)

Permute the dimensions of an input variable without copy.

Parameters:

• x (Variable or numpy.ndarray or cupy.ndarray) – Input variable to be transposed. A \((s_1, s_2, ..., s_N)\) -shaped float array.
• axes (tuple of ints) – By default, reverse the dimensions, otherwise permute the axes according to the values given.

Returns:

Variable whose axes are permuted.

Return type:

Variable

Example

>>> x = np.array([[[0, 1, 2], [3, 4, 5]]], 'f')
>>> x.shape
(1, 2, 3)
>>> y = F.transpose(x).data  # reverse the dimensions
>>> y.shape
(3, 2, 1)
>>> y
array([[[ 0.],
        [ 3.]],

       [[ 1.],
        [ 4.]],

       [[ 2.],
        [ 5.]]], dtype=float32)
>>> y = F.transpose(x, axes=(1, 0, 2)).data # swap 1st and 2nd axis
>>> y.shape
(2, 1, 3)
>>> y
array([[[ 0.,  1.,  2.]],

       [[ 3.,  4.,  5.]]], dtype=float32)