Model State#
The lenskit.state module provides support code for managing the state of
components built on machine learning models, such as extrating learned parameters
and saving model checkpoints.
These features are modeled after the PyTorch state_dict design: components
with state should implement ParameterContainer and expose their state
as dictionaries. This state can then be saved and loaded, including in
pickle-safe formats such as zarr or safetensors.
Protocol for components with learned parameters to enable saving, reloading, checkpointing, etc. |