lenskit.parallel.serialize

Serialization utilities for parallel processing.

Functions

shm_deserialize(data)	Deserialize SHM-pickled data.
shm_serialize(obj[, manager])	Serialize an object for processing in a subclass with shared memory when feasible (including CUDA).

Classes

ModelData(pickle, buffers)	Serialized model data (with shared memory handles).
ModelPickler(file[, protocol, manager, ...])

class lenskit.parallel.serialize.ModelData(pickle, buffers)

Bases: NamedTuple

Serialized model data (with shared memory handles).

Parameters:

• pickle (bytes)

• buffers (list[tuple[SharedMemory | None, int]])

pickle: bytes

Alias for field number 0

buffers: list[tuple[SharedMemory | None, int]]

Alias for field number 1

class lenskit.parallel.serialize.ModelPickler(file, protocol=5, manager=None, *, fix_imports=False)

Bases: Pickler

Parameters:

• protocol (int | None)

• manager (SharedMemoryManager | None)

• fix_imports (bool)

lenskit.parallel.serialize.shm_serialize(obj, manager=None)

Serialize an object for processing in a subclass with shared memory when feasible (including CUDA).

Parameters:

• obj (Any)

• manager (SharedMemoryManager | None)

Return type:

ModelData

lenskit.parallel.serialize.shm_deserialize(data)

Deserialize SHM-pickled data.

Parameters:

data (ModelData)

Return type:

Any