# {py:mod}`dockyard_rl.models.policy.workers.base_policy_worker` ```{py:module} dockyard_rl.models.policy.workers.base_policy_worker ``` ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker :allowtitles: ``` ## Module Contents ### Classes ````{list-table} :class: autosummary longtable :align: left * - {py:obj}`AbstractPolicyWorker ` - ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker :summary: ``` ```` ### API `````{py:class} AbstractPolicyWorker :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker ``` ````{py:method} init_collective(ip: str, port: int, world_size: int, *, train_world_size: int) -> None :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.init_collective ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.init_collective ``` ```` ````{py:method} is_alive() -> bool :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.is_alive ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.is_alive ``` ```` ````{py:method} reset_peak_memory_stats() -> None :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.reset_peak_memory_stats ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.reset_peak_memory_stats ``` ```` ````{py:method} get_gpu_info() -> dict[str, typing.Any] :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_gpu_info ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_gpu_info ``` ```` ````{py:method} report_device_id() -> str :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.report_device_id ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.report_device_id ``` ```` ````{py:method} get_zmq_address() -> str :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_zmq_address ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_zmq_address ``` ```` ````{py:method} maybe_init_zmq() -> None :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.maybe_init_zmq ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.maybe_init_zmq ``` ```` ````{py:method} get_free_memory_bytes() -> int :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_free_memory_bytes ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_free_memory_bytes ``` ```` ````{py:method} get_total_memory_bytes() -> int :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_total_memory_bytes ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_total_memory_bytes ``` ```` ````{py:method} shutdown() -> bool :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.shutdown ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.shutdown ``` ```` ````{py:method} start_gpu_profiling() -> None :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.start_gpu_profiling ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.start_gpu_profiling ``` ```` ````{py:method} stop_gpu_profiling() -> None :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.stop_gpu_profiling ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.stop_gpu_profiling ``` ```` ````{py:method} report_node_ip_and_gpu_id() -> tuple[str, int | str] :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.report_node_ip_and_gpu_id ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.report_node_ip_and_gpu_id ``` ```` ````{py:method} get_reference_policy_logprobs(*, data: BatchedDataDict[Any], micro_batch_size: typing.Optional[int] = None) -> BatchedDataDict[ReferenceLogprobOutputSpec] :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_reference_policy_logprobs ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.get_reference_policy_logprobs ``` ```` ````{py:method} finish_training(*args: typing.Any, **kwargs: typing.Any) -> None :canonical: dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.finish_training ```{autodoc2-docstring} dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker.finish_training ``` ```` `````