torchgear.tensor

exception torchgear.tensor.ShapeError[source]

Bases: RuntimeError

Raised when the shape of a tensor is invalid for some functions.

torchgear.tensor.as1d(x: Tensor, ndim: int = 1, dim: int = -1) → Tensor[source]

Transforms x into a ndim-D tensor but arranges its all elements into dimension dim.

>>> x = torch.arange(4)
>>> as1d(x)
tensor([0, 1, 2, 3])
Parameters:

  • x (Tensor) – A tensor with any shape.

  • ndim (int) – Number of dimensions of resulted tensor.

  • dim – The dimension to place elements in.

Returns:

The transformed tensor.

Return type:

Tensor

torchgear.tensor.broadcastable(*tensors: Tensor) → bool[source]
torchgear.tensor.broadcastable(*shapes: Sequence[int]) → bool

Check whether some tensors or some tensor shapes are broadcastable.

Parameters:

tensors_or_shapes – Some tensors or some tensor shapes.

Returns:

Whether they are broadcastable.

Return type:

bool