Cannot call ray.get_actor in new multiprocess process

1

If I create an actor in the driver process, and try to reference it in a new process from multiprocessing, ray hangs. Is there a workaround that works with multiprocessing?

import ray
import torch.multiprocessing as mp


@ray.remote
class TestActor:
    def __init__(self):
        pass

    def dummy(self):
        return "asdf"


def mp_fn():
    print(1)
    actor = ray.get_actor("dummy")
    print(2)
    print(ray.get(actor.dummy.remote()))
    print(3)


if __name__ == "__main__":
    ray.init()
    actor = TestActor.options(name="dummy").remote()
    p = mp.Process(
        target=mp_fn,
        args=(),
    )
    p.start()