interval

dnois.utils.interval(n: int, spacing: float | Tensor = None, center: float | Tensor = None, symmetric: bool = False, **kwargs) → Tensor

Create a 1D evenly spaced grid.

>>> interval(3, 0.1)
tensor([-0.10, 0.00, 0.10])
>>> interval(3, torch.tensor([0.1, 0.2]))
tensor([[-0.10, 0.00, 0.10],
        [-0.20, 0.00, 0.20]])
>>> interval(3, torch.tensor([0.1, 0.2]), torch.tensor([-1, 1]))
tensor([[-1.10, -1.00, -0.90],
        [ 0.80,  1.00,  1.20]])
>>> interval(4)
tensor([-2., -1., 0., 1.])
>>> interval(4, symmetric=True)
tensor([-1.50, -0.50, 0.50, 1.50])
Parameters:

  • n (int) – Number of grid points.

  • spacing (float or Tensor.) – Spacing between grid points. If a tensor with shape (...), the returned tensor will have shape (..., n). Default: 1.

  • center (float or Tensor.) – Center of resulted grid points. If a tensor with shape (...), the returned tensor will have shape (..., n). Default: 0.

  • symmetric (bool) – If True, grid points are symmetric w.r.t. center. Otherwise, n // 2 points are smaller, n // 2 - 1 points are larger and one point is center value. Only matters when n is even. Default: False.

  • kwargs – Tensor creation arguments passes to torch.linspace() like device.

Returns:

A tensor of shape (n,) if spacing and center are both scalars, otherwise of shape (*<broadcast shape of spacing and center>, n).

Return type:

Tensor