torchrl.trainers.algorithms.configs.data.LazyTensorStorageConfig

class torchrl.trainers.algorithms.configs.data.LazyTensorStorageConfig(_partial_: bool = False, _target_: str = 'torchrl.data.replay_buffers.LazyTensorStorage', max_size: int | None = None, device: Any = 'cpu', ndim: int = 1, compilable: bool = False, consolidated: bool = False, shared_init: bool = False, cleanup_memmap: bool = True)

Hydra configuration for LazyTensorStorage.

Every kwarg accepted by LazyTensorStorage.__init__ is exposed as a field here.