gfn.containers.replay_buffer_manager¶
Classes¶
Module Contents¶
- class gfn.containers.replay_buffer_manager.ReplayBufferManager(env, rank, num_training_ranks, scoring_function=None, diverse_replay_buffer=False, capacity=10000, remote_manager_rank=None)¶
- Parameters:
env (gfn.env.Env)
rank (int)
num_training_ranks (int)
scoring_function (Optional[Callable[[gfn.containers.replay_buffer.ContainerUnion], dict[str, float]]])
diverse_replay_buffer (bool)
capacity (int)
remote_manager_rank (int | None)
- abstract _compute_metadata()¶
- Return type:
dict
- _recv_object()¶
- capacity = 10000¶
- default_scoring_function(obj)¶
Default score function if none provided, placeholder.
- Return type:
dict[str, float]
- diverse_replay_buffer = False¶
- exit_counter = 0¶
- static get_metadata(manager_rank)¶
Sends a get metadata signal to the replay buffer manager.
- Parameters:
manager_rank (int)
- Return type:
dict
- is_running = True¶
- num_training_ranks¶
- rank¶
- remote_manager_rank = None¶
- run()¶
Runs on remote buffer manager ranks. Waits for training data, computes reward, sends back.
- scoring_function¶
- static send_termination_signal(manager_rank)¶
Sends a termination signal to the replay buffer manager.
- Parameters:
manager_rank (int)