reference/generated/torchrl.trainers.algorithms.configs.data.ReplayBufferConfig

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

torchrl.trainers.algorithms.configs.data.ReplayBufferConfig

class torchrl.trainers.algorithms.configs.data.ReplayBufferConfig(_partial_: bool = False, _target_: str = 'torchrl.data.replay_buffers.ReplayBuffer', storage: ~typing.Any = None, sampler: ~typing.Any = None, sample_unit: ~typing.Any = None, writer: ~typing.Any = None, collate_fn: ~typing.Any = None, pin_memory: bool = False, prefetch: int | None = None, transform: ~typing.Any = None, transform_factory: ~typing.Any = None, batch_size: int | None = None, dim_extend: int | None = None, checkpointer: ~typing.Any = None, generator: ~typing.Any = None, consume_after_n_samples: int | None = None, shared: bool = False, compilable: bool | None = None, delayed_init: bool | None = None, service_backend: str = 'direct', service_backend_options: dict[str, ~typing.Any] = <factory>)

Hydra configuration for ReplayBuffer.

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