MCSConfig#

class nvmolkit.mcs.MCSConfig(
batchSize: int = 0,
workerThreads: int = -1,
preprocessingThreads: int = -1,
executorsPerRunner: int = -1,
gpuIds: Sequence[int] | None = None,
)#

Configuration for GPU MCS execution.

Parameters:

  • batchSize – Optional GPU batch chunk size. 0 processes each nonempty size tier as one chunk.

  • workerThreads – GPU runner threads per GPU. -1 autoselects.

  • preprocessingThreads – CPU threads for pair preprocessing. -1 autoselects.

  • executorsPerRunner – Number of asynchronous GPU executor streams used for chunked fMCS tier dispatch. -1 autoselects.

  • gpuIds – GPU device IDs to use. None or empty uses the current device only, not all available devices.

__init__(
batchSize: int = 0,
workerThreads: int = -1,
preprocessingThreads: int = -1,
executorsPerRunner: int = -1,
gpuIds: Sequence[int] | None = None,
) → None#

Initialize GPU execution settings.

to_dict() → dict[str, Any]#

Return a JSON-serializable dictionary of this object’s fields.

to_kwargs() → dict[str, Any]#

Return keyword arguments accepted by findMCS().

classmethod from_dict(
data: dict[str, Any],
) → MCSConfig#

Create an MCSConfig from a dictionary produced by to_dict().