Parameter

A tensor a module registers as a learned weight. Assigning one to a module attribute files it under that name, and it takes part in every tensor operation unchanged (isinstance(p, Tensor) holds). Parameter(data) wraps data without copying: both share one storage. A layer's .weight is a Parameter view of its slot: after the layer packs, a view held across the forward answers its metadata (dtype, shape, device, numel) without a copy, and a value read through the layer (layer.weight) reconstructs the weight once.

data (property)

The plain tensor handle behind this parameter (shared storage). Assigning rebinds the parameter to another tensor's storage in place, so the object a module attribute holds is unchanged by a move or a cast.

__init__

__init____init__(self, data: clika_runtime._core.Tensor) -> None

init(self, data: clika_runtime._core.Tensor) -> None

Wrap data without copying; the parameter shares its storage.