one_hot#

ivy.one_hot(indices, depth, /, *, on_value=None, off_value=None, axis=None, dtype=None, device=None, out=None)[source]#

Return a one-hot array. The locations represented by indices in the parameter indices take value on_value, while all other locations take value off_value.

Parameters:

  • indices (Union[Array, NativeArray]) – Indices for where the ones should be scattered [batch_shape, dim]

  • depth (int) – Scalar defining the depth of the one-hot dimension.

  • on_value (Optional[Number]) – Scalar defining the value to fill in output when indices[j] == i. (default: None) Default: 1.

  • off_value (Optional[Number]) – Scalar defining the value to fill in output when indices[j] != i. (default: None) Default: 0.

  • axis (Optional[int]) – Axis to scatter on. The default is -1, a new inner-most axis is created. (default: None)

  • dtype (Optional[Union[Dtype, NativeDtype]]) – The data type of the output tensor. (default: None)

  • device (Optional[Union[Device, NativeDevice]]) – device on which to create the array ‘cuda:0’, ‘cuda:1’, ‘cpu’ etc. Same as x if (default: None) None.

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

Return type:

Array

Returns:

ret – Tensor of zeros with the same shape and type as a, unless dtype provided which overrides.

Examples

With ivy.Array inputs:

>>> x = ivy.array([3, 1])
>>> y = 5
>>> z = x.one_hot(5)
>>> print(z)
ivy.array([[0., 0., 0., 1., 0.],
...    [0., 1., 0., 0., 0.]])

>>> x = ivy.array([0])
>>> y = 5
>>> ivy.one_hot(x, y)
ivy.array([[1., 0., 0., 0., 0.]])

>>> x = ivy.array([0])
>>> y = 5
>>> ivy.one_hot(x, 5, out=z)
ivy.array([[1., 0., 0., 0., 0.]])
>>> print(z)
ivy.array([[1., 0., 0., 0., 0.]])

With ivy.Container input:

>>> x = ivy.Container(a=ivy.array([1, 2]),         b=ivy.array([3, 1]), c=ivy.array([2, 3]))
>>> y = 5
>>> z = x.one_hot(y)
>>> print(z)
{
    a: ivy.array([[0., 1., 0., 0., 0.],
                [0., 0., 1., 0., 0.]]),
    b: ivy.array([[0., 0., 0., 1., 0.],
                [0., 1., 0., 0., 0.]]),
    c: ivy.array([[0., 0., 1., 0., 0.],
                [0., 0., 0., 1., 0.]])
}

>>> x = ivy.Container(a=ivy.array([2]),         b=ivy.array([]), c=ivy.native_array([4]))
>>> y = 7
>>> z = x.one_hot(y)
>>> print(z)
{
    a: ivy.array([[0., 0., 1., 0., 0., 0., 0.]]),
    b: ivy.array([], shape=(0, 7)),
    c: ivy.array([[0., 0., 0., 0., 1., 0., 0.]])
}
Array.one_hot(self, depth, /, *, on_value=None, off_value=None, axis=None, dtype=None, device=None, out=None)#

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

Parameters:

  • self (Array) – input array containing the indices for which the ones should be scattered

  • depth (int) – Scalar defining the depth of the one-hot dimension.

  • on_value (Optional[Number]) – Value to fill in output when indices[j] == i. Default 1. (default: None)

  • off_value (Optional[Number]) – Value to fill in output when indices[j] != i. Default 0. (default: None)

  • axis (Optional[int]) – The axis to scatter on. The default is -1 which is the last axis. (default: None)

  • dtype (Optional[Union[Dtype, NativeDtype]]) – The data type of the output array. If None, the data type of the on_value is (default: None) used, or if that is None, the data type of the off_value is used. Default float32.

  • device (Optional[Union[Device, NativeDevice]]) – device on which to create the array ‘cuda:0’, ‘cuda:1’, ‘cpu’ etc. (default: None) Same as x if None.

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

Return type:

Array

Returns:

ret – Tensor of zeros with the same shape and type as a, unless dtype provided which overrides.

Examples

With ivy.Array inputs:

>>> x = ivy.array([3, 1])
>>> y = 5
>>> z = x.one_hot(5)
>>> print(z)
ivy.array([[0., 0., 0., 1., 0.],
...    [0., 1., 0., 0., 0.]])

>>> x = ivy.array([0])
>>> y = 5
>>> ivy.one_hot(x, y)
ivy.array([[1., 0., 0., 0., 0.]])

>>> x = ivy.array([0])
>>> y = 5
>>> ivy.one_hot(x, 5, out=z)
ivy.array([[1., 0., 0., 0., 0.]])
>>> print(z)
ivy.array([[1., 0., 0., 0., 0.]])
Container.one_hot(self, depth, /, key_chains=None, to_apply=True, prune_unapplied=False, map_sequences=False, *, on_value=None, off_value=None, axis=None, dtype=None, device=None, out=None)#

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

Parameters:

  • self (Container) – Indices for where the ones should be scattered [batch_shape, dim]

  • depth (int) – Scalar defining the depth of the one-hot dimension.

  • on_value (Optional[Number]) – Value to fill in output when indices[j] == i. If None, defaults to 1. (default: None)

  • off_value (Optional[Number]) – Value to fill in output when indices[j] != i. If None, defaults to 0. (default: None)

  • axis (Optional[int]) – Axis to scatter on. The default is -1, a new inner-most axis is created. (default: None)

  • dtype (Optional[Union[Dtype, NativeDtype]]) – The dtype of the returned tensor. If None, defaults to the on_value dtype (default: None) or the off_value dtype. If both are None, defaults to float32.

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

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

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

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

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

Return type:

Container

Returns:

ret – container with tensors of zeros with the same shape and type as the inputs, unless dtype provided which overrides.

Examples

With ivy.Container input:

>>> x = ivy.Container(a=ivy.array([1, 2]),              b=ivy.array([3, 1]), c=ivy.array([2, 3]))
>>> y = 5
>>> z = x.one_hot(y)
>>> print(z)
{
    a: ivy.array([[0., 1., 0., 0., 0.],
                [0., 0., 1., 0., 0.]]),
    b: ivy.array([[0., 0., 0., 1., 0.],
                [0., 1., 0., 0., 0.]]),
    c: ivy.array([[0., 0., 1., 0., 0.],
                [0., 0., 0., 1., 0.]])
}

>>> x = ivy.Container(a=ivy.array([1, 2]),              b=ivy.array([]), c=ivy.native_array([4]))
>>> y = 5
>>> z = x.one_hot(y)
>>> print(z)
{
    a: ivy.array([[0., 1., 0., 0., 0.],
                [0., 0., 1., 0., 0.]]),
    b: ivy.array([], shape=(0, 5)),
    c: ivy.array([[0., 0., 0., 0., 1.]])
}