jax.numpy.rot90

jax.numpy.rot90(m, k=1, axes=(0, 1))

Rotate an array by 90 degrees counterclockwise in the plane specified by axes.

JAX implementation of numpy.rot90().

Parameters:

• m (ArrayLike) – input array. Must have m.ndim >= 2.

• k (int) – int, optional, default=1. Specifies the number of times the array is rotated. For negative values of k, the array is rotated in clockwise direction.

• axes (tuple[int, int]) – tuple of 2 integers, optional, default= (0, 1). The axes define the plane in which the array is rotated. Both the axes must be different.

Returns:

An array containing the copy of the input, m rotated by 90 degrees.

Return type:

Array

See also

• jax.numpy.flip(): reverse the order along the given axis

• jax.numpy.fliplr(): reverse the order along axis 1 (left/right)

• jax.numpy.flipud(): reverse the order along axis 0 (up/down)

Examples

>>> m = jnp.array([[1, 2, 3],
...                [4, 5, 6]])
>>> jnp.rot90(m)
Array([[3, 6],
       [2, 5],
       [1, 4]], dtype=int32)
>>> jnp.rot90(m, k=2)
Array([[6, 5, 4],
       [3, 2, 1]], dtype=int32)

jnp.rot90(m, k=1, axes=(1, 0)) is equivalent to jnp.rot90(m, k=-1, axes(0,1)).

>>> jnp.rot90(m, axes=(1, 0))
Array([[4, 1],
       [5, 2],
       [6, 3]], dtype=int32)
>>> jnp.rot90(m, k=-1, axes=(0, 1))
Array([[4, 1],
       [5, 2],
       [6, 3]], dtype=int32)

when input array has ndim>2:

>>> m1 = jnp.array([[[1, 2, 3],
...                  [4, 5, 6]],
...                 [[7, 8, 9],
...                  [10, 11, 12]]])
>>> jnp.rot90(m1, k=1, axes=(2, 1))
Array([[[ 4,  1],
        [ 5,  2],
        [ 6,  3]],

       [[10,  7],
        [11,  8],
        [12,  9]]], dtype=int32)