lucid.diag

lucid.diag(v: Tensor | list | ndarray | array, k: int = 0, /, dtype: type[bool | int | float | complex] | Numeric | None = None, requires_grad: bool = False, keep_grad: bool = False, device: Literal['cpu', 'gpu'] | None = None) → Tensor

The diag function extracts a diagonal or constructs a diagonal tensor from an input tensor or array.

Function Signature

def diag(
    v: Tensor | _ArrayLike,
    k: int = 0,
    dtype: Any = np.float32,
    requires_grad: bool = False,
    keep_grad: bool = False,
    device: _DeviceType | None = None,
) -> Tensor

Parameters

• v (Tensor | _ArrayLike):

  The input tensor or array. If it is a 1D array, a 2D tensor with the specified diagonal is returned. If it is a 2D array, the specified diagonal is extracted.

• k (int, optional):

  The index of the diagonal. A value of 0 refers to the main diagonal, k > 0 refers to a diagonal above the main diagonal, and k < 0 refers to a diagonal below it. Defaults to 0.

• dtype (Any, optional):

  The data type of the output tensor. Defaults to np.float32.

• requires_grad (bool, optional):

  If True, the resulting tensor will be part of the computation graph and capable of tracking gradients. Defaults to False.

• keep_grad (bool, optional):

  If True, the gradient history will be preserved even if the tensor does not

  require gradients. Defaults to False.

Returns

• Tensor:

  A tensor representing the extracted diagonal or the constructed diagonal tensor.

Example

>>> import lucid
>>> a = Tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
>>> d = lucid.diag(a)
>>> print(d)
Tensor([1, 5, 9])

Constructing a diagonal tensor from a 1D array:

>>> v = Tensor([1, 2, 3])
>>> d = lucid.diag(v)
>>> print(d)
Tensor([[1. 0. 0.]
        [0. 2. 0.]
        [0. 0. 3.]])

Note

• Supports both diagonal extraction and construction depending on the input’s dimensionality.

• Use the k parameter to specify diagonals above or below the main diagonal.