caikit.runtime.service_factory

This module is responsible for creating service objects for the runtime to consume

Attributes

log
error

Classes

ServicePackage	Container with references to:
ServicePackageFactory	Factory responsible for yielding the correct concrete ServicePackage implementation

Functions

get_inference_request(...)	Helper function to return the inference request DataModel for the Module or Task Class
get_train_request(...)	Helper function to return the train request DataModel for the Module Class
get_train_params(...)	Helper function to return the train parameters DataModel for the Module Class

Module Contents

caikit.runtime.service_factory.log

caikit.runtime.service_factory.error

class caikit.runtime.service_factory.ServicePackage

Container with references to: - A service Class - A service Descriptor - A grpc servicer registration function - A client stub - A client messages module

service: Type[google.protobuf.service.Service]

descriptor: google.protobuf.descriptor.ServiceDescriptor

registration_function: Callable[[google.protobuf.service.Service, grpc.Server], None]

stub_class: Type

messages: types.ModuleType

caikit_rpcs: Dict[str, caikit.runtime.service_generation.rpcs.CaikitRPCBase]

class caikit.runtime.service_factory.ServicePackageFactory

Factory responsible for yielding the correct concrete ServicePackage implementation

ServiceType

classmethod get_service_package(service_type: ServiceType, write_modules_file: bool = False) → ServicePackage

Public factory API. Returns a service package of the requested type, from the configured source.

Args:

service_type (ServicePackageFactory.ServiceType): The type of service to build,

to match the servicer implementation that will handle it. e.g. the GlobalPredictServicer expects an “INFERENCE” service

write_modules_file (bool): if set, write out a modules.json file to list the

included modules in this service generation. See config to customize file name at runtime.service_generation.backwards_compatibility.current_modules_path

Returns:

ServicePackage: A container with properties referencing everything you need to bind a

concrete Servicer implementation to a protobufs Service and grpc Server

static _check_backwards_compatibility(caikit_config: aconfig.Config, clean_modules: Set[Type[caikit.core.ModuleBase]])

static _get_and_filter_modules(caikit_config: aconfig.Config, lib: str, write_modules_file: bool) → Set[Type[caikit.core.ModuleBase]]

caikit.runtime.service_factory.get_inference_request(task_or_module_class: Type[caikit.core.ModuleBase | caikit.core.task.TaskBase], input_streaming: bool = False, output_streaming: bool = False) → Type[caikit.core.data_model.base.DataBase]

Helper function to return the inference request DataModel for the Module or Task Class

caikit.runtime.service_factory.get_train_request(module_class: Type[caikit.core.ModuleBase]) → Type[caikit.core.data_model.base.DataBase]

Helper function to return the train request DataModel for the Module Class

caikit.runtime.service_factory.get_train_params(module_class: Type[caikit.core.ModuleBase]) → Type[caikit.core.data_model.base.DataBase]

Helper function to return the train parameters DataModel for the Module Class