ray.experimental.register_tensor_transport#

ray.experimental.register_tensor_transport(transport_name: str, devices: List[str], transport_manager_class: type[TensorTransportManager])[source]#

Register a new tensor transport for use in Ray. Note that this needs to be called before you create the actors that will use the transport. The actors also need to be created in the same process from which you call this function.

Parameters:

  • transport_name – The name of the transport protocol.

  • devices – List of PyTorch device types supported by this transport (e.g., [“cuda”, “cpu”]).

  • transport_manager_class – A class that implements TensorTransportManager.

Raises:

ValueError – If transport_manager_class is not a subclass of TensorTransportManager.

PublicAPI (alpha): This API is in alpha and may change before becoming stable.