lucid.load

lucid.load(path: Path | str) → Tensor | Module | OrderedDict

The load function deserializes and loads a previously saved Lucid object from disk. It supports loading Tensor objects from .lct files and OrderedDict state_dicts from .lcd files.

Function Signature

def load(path: Path | str) -> Tensor | OrderedDict

Parameters

  • path (Path | str): Path to the file containing the saved object. The file must have a .lct or .lcd suffix.

Returns

  • Tensor: If the file contains a saved Tensor (from .lct file).

  • OrderedDict: If the file contains a saved state_dict (from .lcd file).

Behavior

  • If the file has a .lct extension, the function loads the raw NumPy array and wraps it as a Tensor.

  • If the file has a .lcd extension, the function returns the OrderedDict representing the model’s state_dict.

Warning

The function assumes that the file was created using lucid.save. Files not conforming to Lucid’s format will raise an error.

Examples

from lucid import load

# Load a tensor
t = load("tensor.lct")
print(t)  # Tensor([[1.0, 2.0]], grad=None)

# Load a state_dict
state = load("weights.lcd")
print(state.keys())  # dict_keys([...])

Note

This function relies on pickle for deserialization. Ensure that the file was created in a compatible Python environment.