Skip to content

convert

Utility functions for converting between different data types.

to_numpy

Convert an iterable (including a tensor or tensor subclasses like torch_geometric.EdgeIndex) to numpy.

Parameters:

Name Type Description Default
input_iterable torch.Tensor | numpy.ndarray | list | tuple

Tensor, tensor subclass, numpy array or list.

 required

Returns:

Type Description
numpy.ndarray

Numpy array.
Source code in src/pathpyG/utils/convert.py 
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
def to_numpy(input_iterable: torch.Tensor | np.ndarray | list | tuple) -> np.ndarray:
    """Convert an iterable (including a tensor or tensor subclasses like [`torch_geometric.EdgeIndex`][torch_geometric.EdgeIndex]) to numpy.

    Args:
        input_iterable: [Tensor][torch.Tensor], tensor subclass, [numpy array][numpy.ndarray] or list.

    Returns:
        Numpy array.
    """
    if isinstance(input_iterable, (EdgeIndex, Index)):
        return input_iterable.as_tensor().cpu().numpy()
    elif isinstance(input_iterable, torch.Tensor):
        return input_iterable.cpu().numpy()
    elif isinstance(input_iterable, (list, tuple)):
        return np.array(input_iterable)
    elif isinstance(input_iterable, np.ndarray):
        return input_iterable