ThreadingTransport

class torchrl.modules.inference_server.ThreadingTransport

In-process transport for actors that are threads.

Uses a shared list protected by a threading.Condition as the request queue and Future objects for response routing.

This is the simplest backend and is appropriate when all actors live in the same process (e.g. running in a ThreadPoolExecutor).

drain(max_items: int) → tuple[list[TensorDictBase], list[Future]]

Dequeue up to max_items pending requests.

resolve(callback: Future, result: TensorDictBase) → None

Set the result on the actor’s Future.

resolve_exception(callback: Future, exc: BaseException) → None

Set an exception on the actor’s Future.

submit(td: TensorDictBase) → Future[TensorDictBase]

Enqueue a request and return a Future for the result.

wait_for_work(timeout: float) → None

Block until at least one request is enqueued or timeout elapses.