verl/single_controller/base/worker_group.py

• ResourcePool · class · L26-L57 — class ResourcePool
• init · method · L28-L33 — def init(self, process_on_nodes=None, max_collocate_count: int = 10, n_gpus_per_node=8) -> None
• add_node · method · L35-L36 — def add_node(self, process_count)
• world_size · method · L39-L40 — def world_size(self)
• call · method · L42-L43 — def call(self) -> Any
• store · method · L46-L47 — def store(self)
• local_world_size_list · method · L49-L53 — def local_world_size_list(self) -> List[int]
• local_rank_list · method · L55-L57 — def local_rank_list(self) -> List[int]
• ClassWithInitArgs · class · L60-L78 — class ClassWithInitArgs
• init · method · L66-L69 — def init(self, cls, args, *kwargs) -> None
• call · method · L77-L78 — def call(self) -> Any
• check_workers_alive · function · L81-L88 — def check_workers_alive(workers: List, is_alive: Callable, gap_time: float = 1) -> None
• WorkerGroup · class · L91-L196 — class WorkerGroup
• init · method · L93-L108 — def init(self, resource_pool: ResourcePool, **kwargs) -> None
• _is_worker_alive · method · L110-L111 — def _is_worker_alive(self, worker)
• _block_until_all_workers_alive · method · L113-L119 — def _block_until_all_workers_alive(self) -> None
• start_worker_aliveness_check · method · L121-L127 — def start_worker_aliveness_check(self, every_n_seconds=1) -> None: # before starting checking worker aliveness, make sure all workers are already alive
• world_size · method · L130-L131 — def world_size(self)
• _bind_worker_method · method · L136-L196 — def _bind_worker_method(self, user_defined_cls, func_generator)