ray.actor.ActorMethod.__init__#

ActorMethod.__init__(actor, method_name, num_returns: int | Literal['streaming'] | None, max_task_retries: int, retry_exceptions: bool | list | tuple, is_generator: bool, generator_backpressure_num_objects: int, enable_task_events: bool, decorator=None, signature: List[Parameter] | None = None, hardref=False, tensor_transport: Literal[ray.core.generated.common_pb2.OBJECT_STORE, ray.core.generated.common_pb2.NCCL, ray.core.generated.common_pb2.GLOO] | None = None)[source]#

Initialize an ActorMethod.

Parameters:
  • actor – The actor instance this method belongs to.

  • method_name – The name of the actor method.

  • num_returns – The default number of return values that the method invocation should return. If None is given, it uses DEFAULT_ACTOR_METHOD_NUM_RETURN_VALS for a normal actor task and “streaming” for a generator task (when is_generator is True).

  • max_task_retries – Number of retries on method failure.

  • retry_exceptions – Boolean of whether you want to retry all user-raised exceptions, or a list of allowlist exceptions to retry.

  • is_generator – True if a given method is a Python generator.

  • generator_backpressure_num_objects – Generator-only config. If a number of unconsumed objects reach this threshold, a actor task stop pausing.

  • enable_task_events – True if task events is enabled, i.e., task events from the actor should be reported. Defaults to True.

  • decorator – An optional decorator that should be applied to the actor method invocation.

  • signature – The signature of the actor method. It is None only when cross language feature is used.

  • hardref – Whether to keep a hard reference to the actor.

  • tensor_transport – The tensor transport protocol to use for the actor method. The valid values are OBJECT_STORE (default), NCCL, or GLOO, and they are case-insensitive.