Vectorized Buffer Helpers

Note

This package is experimental.

Experimental vectorized rollout buffer wrappers.

class DummyVecRolloutBuffer(buffer_size: int, observation_space: Space, action_space: Space, device: device | str = 'auto', n_envs: int = 1, use_threads: bool = False, buffers: Iterable[BaseBuffer] | None = None, **kwargs)

Bases: RolloutBuffer

Forward rollout API calls to a list of underlying buffers.

Create a facade over multiple rollout buffers.

Parameters:

• buffer_size – Steps per environment for each underlying buffer.

• observation_space – Shared observation space.

• action_space – Shared action space.

• device – Torch device passed to the base rollout buffer.

• n_envs – Number of parallel environments for the facade.

• use_threads – Run delegated calls in a thread pool when True.

• buffers – Underlying buffers to invoke; defaults to an empty list.

• **kwargs – Additional arguments forwarded to RolloutBuffer.

property full

Return whether every underlying buffer is full.

Implementation modules

Experimental rollout buffer that delegates to multiple underlying buffers.

class DummyVecRolloutBuffer(buffer_size: int, observation_space: Space, action_space: Space, device: device | str = 'auto', n_envs: int = 1, use_threads: bool = False, buffers: Iterable[BaseBuffer] | None = None, **kwargs)

Bases: RolloutBuffer

Forward rollout API calls to a list of underlying buffers.

Create a facade over multiple rollout buffers.

Parameters:

• buffer_size – Steps per environment for each underlying buffer.

• observation_space – Shared observation space.

• action_space – Shared action space.

• device – Torch device passed to the base rollout buffer.

• n_envs – Number of parallel environments for the facade.

• use_threads – Run delegated calls in a thread pool when True.

• buffers – Underlying buffers to invoke; defaults to an empty list.

• **kwargs – Additional arguments forwarded to RolloutBuffer.

property full

Return whether every underlying buffer is full.