einops_rearrange#

ivy.einops_rearrange(x, pattern, /, *, out=None, **axes_lengths)[source]#

Perform einops rearrange operation on input array x.

Parameters:

  • x (Union[Array, NativeArray]) – Input array to be re-arranged.

  • pattern (str) – Rearrangement pattern.

  • axes_lengths (Dict[str, int]) – Any additional specifications for dimensions.

  • out (Optional[Array], default: None) – optional output array, for writing the result to. It must have a shape that the inputs broadcast to.

Return type:

Array

Returns:

ret – New array with einops.rearrange having been applied.

Examples

With ivy.Array instance method:

>>> x = ivy.array([[1, 2, 3],
...               [-4, -5, -6]])
>>> y = x.einops_rearrange("height width -> width height")
>>> print(y)
ivy.array([[ 1, -4],
       [ 2, -5],
       [ 3, -6]])

>>> x = ivy.array([[[ 1,  2,  3],
...                  [ 4,  5,  6]],
...               [[ 7,  8,  9],
...                  [10, 11, 12]]])
>>> y = x.einops_rearrange("c h w -> c (h w)")
>>> print(y)
ivy.array([[ 1,  2,  3,  4,  5,  6],
       [ 7,  8,  9, 10, 11, 12]])

>>> x = ivy.array([[1, 2, 3, 4, 5, 6],
...            [7, 8, 9, 10, 11, 12]])
>>> y = ivy.zeros((4,3))
>>> x.einops_rearrange("c (h w) -> (c h) w", out=y, h=2, w=3)
>>> print(y)
ivy.array([[ 1,  2,  3],
       [ 4,  5,  6],
       [ 7,  8,  9],
       [10, 11, 12]])

With ivy.Container input:

>>> x = ivy.Container(a=ivy.array([[-4.47, 0.93, -3.34],
...                            [3.66, 24.29, 3.64]]),
...               b=ivy.array([[4.96, 1.52, -10.67],
...                            [4.36, 13.96, 0.3]]))
>>> y = ivy.einops_rearrange(x, 'a b -> b a')
>>> print(y)
{
    a: ivy.array([[-4.46999979, 3.66000009],
                  [0.93000001, 24.29000092],
                  [-3.33999991, 3.6400001]]),
    b: ivy.array([[4.96000004, 4.36000013],
                  [1.51999998, 13.96000004],
                  [-10.67000008, 0.30000001]])
}

With varying pattern:

Suppose we have a set of 32 images in “h w c” format (height-width-channel) and concatenate images along height (vertical axis), 960 = 32 * 30

>>> images = ivy.asarray([ivy.random_normal(shape=(30, 40, 3)) for _ in range(32)])
>>> x = ivy.einops_rearrange(images, 'b h w c -> (b h) w c')
>>> print(x.shape)
(960, 40, 3)

# Concatenate images along horizontal axis, 1280 = 32 * 40

>>> images = ivy.asarray([ivy.random_normal(shape=(30, 40, 3)) for _ in range(32)])
>>> x = ivy.einops_rearrange(images, 'b h w c -> h (b w) c')
>>> print(x.shape)
(30, 1280, 3)

# Reorder axes to “b c h w” format for deep learning

>>> images = ivy.asarray([ivy.random_normal(shape=(30, 40, 3)) for _ in range(32)])
>>> x = ivy.einops_rearrange(images, 'b h w c -> b c h w')
>>> print(x.shape)
(32, 3, 30, 40)

# Flatten each image into a vector, 3600 = 30 * 40 * 3

>>> images = ivy.asarray([ivy.random_normal(shape=(30, 40, 3)) for _ in range(32)])
>>> x = ivy.einops_rearrange(images, 'b h w c -> b (c h w)')
>>> print(x.shape)
(32, 3600)

# Split each image into 4 smaller (top-left, top-right, bottom-left, bottom-right), # 128 = 32 * 2 * 2

>>> images = ivy.asarray([ivy.random_normal(shape=(30, 40, 3)) for _ in range(32)])
>>> x = ivy.einops_rearrange(images, 'b (h1 h) (w1 w) c -> (b h1 w1) h w c',
... h1=2, w1=2)
>>> print(x.shape)
(128, 15, 20, 3)

# Space-to-depth operation >>> images = ivy.asarray([ivy.random_normal(shape=(30, 40, 3)) for _ in range(32)]) >>> x = ivy.einops_rearrange(images, ‘b (h h1) (w w1) c -> b h w (c h1 w1)’, h1=2, … w1=2) >>> print(x.shape) (32, 15, 20, 12)

Array.einops_rearrange(self, pattern, /, *, out=None, **axes_lengths)[source]#

ivy.Array instance method variant of ivy.einops_rearrange. This method simply wraps the function, and so the docstring for ivy.einops_rearrange also applies to this method with minimal changes.

Parameters:

  • self (Array) – Input array to be re-arranged.

  • pattern (str) – Rearrangement pattern.

  • axes_lengths (Dict[str, int]) – Any additional specifications for dimensions.

  • out (Optional[Array], default: None) – optional output array, for writing the result to. It must have a shape that the inputs broadcast to.

Return type:

Array

Returns:

ret – New array with einops.rearrange having been applied.

Examples

With ivy.Array instance method:

>>> x = ivy.array([[1, 2, 3],
...               [-4, -5, -6]])
>>> y = x.einops_rearrange("height width -> width height")
>>> print(y)
ivy.array([[ 1, -4],
    [ 2, -5],
    [ 3, -6]])

>>> x = ivy.array([[[ 1,  2,  3],
...                  [ 4,  5,  6]],
...               [[ 7,  8,  9],
...                  [10, 11, 12]]])
>>> y = x.einops_rearrange("c h w -> c (h w)")
>>> print(y)
ivy.array([[ 1,  2,  3,  4,  5,  6],
    [ 7,  8,  9, 10, 11, 12]])

>>> x = ivy.array([[1, 2, 3, 4, 5, 6]
...               [7, 8, 9, 10, 11, 12]])
>>> y = x.einops_rearrange("c (h w) -> (c h) w", h=2, w=3)
ivy.array([[ 1,  2,  3],
    [ 4,  5,  6],
    [ 7,  8,  9],
    [10, 11, 12]])
Container.einops_rearrange(self, pattern, /, *, key_chains=None, to_apply=True, prune_unapplied=False, map_sequences=False, out=None, **axes_lengths)[source]#

ivy.Container instance method variant of ivy.einops_rearrange. This method simply wraps the function, and so the docstring for ivy.einops_rearrange also applies to this method with minimal changes.

Parameters:

  • pattern (Union[str, Container]) – Rearrangement pattern.

  • key_chains (Optional[Union[List[str], Dict[str, str], Container]], default: None) – The key-chains to apply or not apply the method to. Default is None.

  • to_apply (Union[bool, Container], default: True) – If True, the method will be applied to key_chains, otherwise key_chains will be skipped. Default is True.

  • prune_unapplied (Union[bool, Container], default: False) – Whether to prune key_chains for which the function was not applied. Default is False.

  • map_sequences (Union[bool, Container], default: False) – Whether to also map method to sequences (lists, tuples). Default is False.

  • axes_lengths (Union[Dict[str, int], Container]) – Any additional specifications for dimensions.

  • **axes_lengths –

Returns:

ivy.Container with each array having einops.rearrange applied.

Examples

>>> x = ivy.Container(a=ivy.array([[1, 2, 3],
...                                [-4, -5, -6]]),
...                 b=ivy.array([[7, 8, 9],
...                              [10, 11, 12]]))
>>> y = x.einops_rearrange("height width -> width height")
>>> print(y)
{
    a: ivy.array([[1, -4],
                [2, -5],
                [3, -6]]),
    b: ivy.array([[7, 10],
                [8, 11],
                [9, 12]])
}

>>> x = ivy.Container(a=ivy.array([[[ 1,  2,  3],
...                  [ 4,  5,  6]],
...               [[ 7,  8,  9],
...                  [10, 11, 12]]]))
>>> y = x.einops_rearrange("c h w -> c (h w)")
>>> print(y)
{
    a: (<class ivy.data_classes.array.array.Array> shape=[2, 6])
}

>>> x = ivy.Container(a=ivy.array([[1, 2, 3, 4, 5, 6],
...               [7, 8, 9, 10, 11, 12]]))
>>> y = x.einops_rearrange("c (h w) -> (c h) w", h=2, w=3)
>>> print(y)
{
    a: (<class ivy.data_classes.array.array.Array> shape=[4, 3])
}