OakEmu-RL v0.1.0

Appearance

DQNAgent ​

python
DQNAgent(model: DQNModel | ThreadSafeModel, device: torch.device | None = None, observation_batch_shape_hint: tuple[int, ...] | None = None) -> None

Agent responsible for greedy action selection and Q-value computation.

Exploration logic (epsilon-greedy, noisy nets) is handled separately by ExplorationStrategy classes in the Actor.

Initialize the DQNAgent.

Parameters:

  • model (DQNModel | ThreadSafeModel) – The DQN model or a ThreadSafeModel container.
  • device (device | None) – The device to run inference on.
  • observation_batch_shape_hint (tuple[int, ...] | None) – Optional maximum batch shape for the reusable CUDA observation buffer.

Source

Methods ​

get_action ​

python
get_action(state: torch.Tensor | np.ndarray, action_masker: ActionMasker | None = None) -> int

Choose the greedy action based on Q-values.

Parameters:

  • state (Tensor | ndarray) – The current environment state.
  • action_masker (ActionMasker | None) – Optional ActionMasker for filtering allowed actions.

Returns:

  • int – The chosen action (greedy).

Source

get_actions ​

python
get_actions(states: np.ndarray, action_masker: ActionMasker | None = None) -> list[int]

Choose greedy actions for a batch of states.

Parameters:

  • states (ndarray) – A batch of state observations with shape (N, *state_shape).
  • action_masker (ActionMasker | None) – Optional ActionMasker for filtering allowed actions.

Returns:

  • list[int] – A list of selected actions (greedy).

Source

get_q_values ​

python
get_q_values(states: torch.Tensor | np.ndarray) -> torch.Tensor

Compute Q-values for the given states.

Parameters:

  • states (Tensor | ndarray) – A batch of states.

Returns:

  • Tensor – The computed Q-values.

Source

Values ​

device ​

python
device = device if device is not None else torch.device('cpu')

Source

model ​

python
model = model

Source