tensor.py

rerun.log.tensor

Tensor = Union[npt.ArrayLike, TorchTensorLike] module-attribute

Type helper for a tensor-like object that can be logged to Rerun.

class TorchTensorLike

Bases: Protocol

Describes what is need from a Torch Tensor to be loggable to Rerun.

def log_tensor(entity_path, tensor, *, names=None, meter=None, ext=None, timeless=False)

Log an n-dimensional tensor.

Parameters:

Name	Type	Description	Default
entity_path	str	Path to the tensor in the space hierarchy.	required
tensor	npt.ArrayLike	A Tensor object.	required
names	Optional[Iterable[Optional[str]]]	Optional names for each dimension of the tensor.	None
meter	Optional[float]	Optional scale of the tensor (e.g. meters per cell).	None
ext	Optional[Dict[str, Any]]	Optional dictionary of extension components. See rerun.log_extension_components	None
timeless	bool	If true, the tensor will be timeless (default: False).	False