mmengine.dataset.worker_init_fn¶ mmengine.dataset.worker_init_fn(worker_id, num_workers, rank, seed)[source]¶ This function will be called on each worker subprocess after seeding and before data loading. Parameters worker_id (int) – Worker id in [0, num_workers - 1]. num_workers (int) – How many subprocesses to use for data loading. rank (int) – Rank of process in distributed environment. If in non-distributed environment, it is a constant number 0. seed (int) – Random seed. Return type None