^jIddlmZddlmZddlmZddlmZddlm Z ddgZ GddeZ Gd d e Z ed Gd de Z y ))Sized)IntEnum)NoReturn)functional_datapipe) IterDataPipeSHARDING_PRIORITIESShardingFilterIterDataPipeceZdZdZdZdZy)rN)__name__ __module__ __qualname__DEFAULT DISTRIBUTEDMULTIPROCESSINGs/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/torch/utils/data/datapipes/iter/sharding.pyrrsGKOrc$eZdZdedededefdZy)_ShardingIterDataPipenum_of_instances instance_idsharding_groupreturnctN)NotImplementedErrorselfrrrs rapply_shardingz$_ShardingIterDataPipe.apply_shardings "!rN)rrrintrrr"rrrrrs-""", "  "rrsharding_filtercheZdZdZ d deddffd ZejfdZd dZ dZ de fd Z xZ S) r ao Wrapper that allows DataPipe to be sharded (functional name: ``sharding_filter``). After ``apply_sharding`` is called, each instance of the DataPipe (on different workers) will have every `n`-th element of the original DataPipe, where `n` equals to the number of instances. Args: source_datapipe: Iterable DataPipe that will be sharded Nsource_datapiperct|||_||_i|_d|_d|_|jyNr r)super__init__r&sharding_group_filtergroupsrr_update_num_of_instances)r!r&r+ __class__s rr*z#ShardingFilterIterDataPipe.__init__,sC .%:"24 ! %%'rc`||k\rtd|d|d|tjk(rycc}wr()sortedr,keysr+reverserr)r!keysorted_sharding_groupsgroup_num_of_instancesgroup_instance_ids rr-z3ShardingFilterIterDataPipe._update_num_of_instancesKsdkk..01" ))1SDrNsc!E<   ' "L"&'IU!6IU(IUr