i#VddlZddlmZddlmZmZddlZddlmZddl m Z ddl m Z ddl mZm Z dZGd d e Zd ed ed efdZGddZe e ddej&deejj&dzd ej&fdZdej&d efdZy)N)Iterable)AnyNoReturn)_State)DistributedDataParallel) _get_registrycontractc eZdZUejed<dfd Zefdejde ejde ejde ddf d Z dd Zdejde ejddfd Zdd Zdd Zdejdeedfdee efdefdZdejdeej.dej.dej.fdZxZS)_ReplicateState _ddp_weakrefreturnNct|tj|_d|_tj|_|j|_g|_d|_ d|_ i|_ g|_ y)NF) super__init__nn ParameterListmodulehas_initialized _param_list _orig_module _param_names_no_sync _init_args _init_kwargs_comm_hook_args)self __class__s a/var/www/html/engine/venv/lib/python3.12/site-packages/torch/distributed/_composable/replicate.pyrz_ReplicateState.__init__sk !#!1!1!3 %*-/-=-=-?!KK')# 26,.*,rignored_modulesignored_paramsprefixcbt|ry||vry|tk7r|dnt}|jdD]D\}}||vs |jj ||j j ||F|j D]\}} |j| |||| y)N.F)recurse)r$)_is_fully_sharded_ROOT_MODULE_PREFIXnamed_parametersrappendrnamed_children_collect_params) rrr"r#r$recurse_prefixnpname child_modules r r-z_ReplicateState._collect_params"s V $  _ $ #&99vhaL?R ++E+: ADAq&  ''*!!((N+;A3)?@ A #)"7"7"9  D,  ()$0 !  r!cPtjdfd}|y)NT) recursivecjJjjijjd_i_y)N)rinitrregister_comm_hook)rsr _lazy_initz-_ReplicateState.lazy_init.._lazy_initCsK??. .. DIIt <$*;*; <  # # % DO "D r!)torch_disable_dynamo)rr9s` r lazy_initz_ReplicateState.lazy_initBs'    . # / #  r!c |jryd|_||_|Dchc]}|jD]}|}}}|jD]-}t |s|j |j/ddlm}||||j|||d|vrV|d;|d} t| tjr| jdk(rd|d<n | g|d<nd|d<|jdt|jfi||_t#j$|j t&j)|j_ycc}}w)NTr)_localize_dtensor)r# device_idcpu device_ids)rr parametersmodulesr(update%torch.distributed.tensor.parallel.ddpr>r- isinstancer:devicetypepoprr_ddpweakrefref replicatestater) rrr"kwargsmr0r# submoduler>r?s r r7z_ReplicateState.initMs@    # %4MallnM!M!MM) >I +%%i&:&:&<= > L&@ V_nE & k".";/ i69>>U;R+/F<(-6;F<('+|$ JJ{ #+D,<,<GG 4;KK 4J  $1;NsEc|jD]!\}}|jj|i|#|jjyN)rrJr8clear)r comm_args comm_kwargss r r8z"_ReplicateState.register_comm_hookwsI&*&:&: D "I{ (DII ( () C{ C D ""$r!c ||_||_yrS)rrrargsrOs r record_init_argsz _ReplicateState.record_init_args|s"r!rY.rOc|js |jr|j|j |j_|jj |i|SrS)rrr<rrJrequire_backward_grad_sync _pre_forward)rrrYrOs r forward_pre_hookz _ReplicateState.forward_pre_hooksL ??d// NN 37==/@ ,%tyy%%t6v66r!inputoutputc8|jj|SrS)rJ _post_forward)rrr_r`s r forward_post_hookz!_ReplicateState.forward_post_hooks yy&&v..r!rN)__name__ __module__ __qualname__rKrL__annotations__rr)rModuleset Parameterstrr-r<r7r8rZtuplerdictr^r:Tensorrc __classcell__)rs@r r r s++ -&*  RYYBLL)     @ (K (KRYY(K  (KT% #7ii7',S#X7@DS#X7 7/ /U\\"/ /  /r!r rYrOrctd)NzGDDP does not support deepcopy. Please use state dict for serialization.)AssertionError)rYrOs r unimplemented_deepcopyrss Q r!c*eZdZdZdeddfdZddZy)DDPcL|jd}|j|g|i|S)z Override ``__new__`` to remove the DDP class and directly construct the original class for cases like indexing into a container module. )__mro____new__)clsrYrOorig_clss r ryz DDP.__new__s.;;q>x:4:6::r!requires_gradient_syncrNc:| tj|_y)a Sets if the module should sync gradients. This can be used to implement gradient accumulation without communication. Args: requires_gradient_sync (bool): Whether to reduce gradients for the module's parameters. N)rMrNr)rr|s r set_requires_gradient_synczDDP.set_requires_gradient_syncs.D)C &r!cdtj|jj||fyrS)rMrNrr+rXs r r8zDDP.register_comm_hooks#--44dF^Dr!rd)rerfrgryboolr~r8r6r!r rurus#; D D$ DEr!ru) state_clsrr"c tjjdd|vr=t|dttj fst dt|dt|r t d|i}n t|}tj|}|j|jd|jd}|?|j}||k7r*dd lm}m}|j||j'||j'|j(|j*||fi||j,}d t.i} td |j0t2|f| } | |_|S) zReplicates a module Args: module (torch.nn.Module): module to replicate Example:: >>> # xdoctest: +REQUIRES(module:torch._C._distributed_c10d) >>> module = nn.Linear(3, 3) >>> replicate(module) ztorch.distributed.replicater?z6Expected device_id to be int or torch.device, but got zGCannot apply `replicate()` on a Module already managed by `fully_shard`T) with_kwargs device_meshr)r>_reconstruct_dtensor __deepcopy__ru)r:_C_log_api_usage_oncerFintrG RuntimeErrorrHr(rjrMrNregister_forward_pre_hookr^get_get_root_meshrEr>rregister_forward_hookrcrZrrsreru) rr"rOrNr root_meshr>rrzdctnew_clss r rMrMsv  HH  !>?f&-U\\/BC{ 3457    U  o. OOF #E $$U%;%;$N**]+K..0   #   , ,-A B  ( (): ;   !8!89E6?=f=   C 1 2CS'#sS9GF Mr!c&t|}|yd|vS)z+Check if module is marked with fully_shard.F fully_shard)r )rregistrys r r(r(s V$H H $$r!rS)rKcollections.abcrtypingrrr:torch.nnr#torch.distributed._composable_statertorch.nn.parallelrr r r)r rsrurirMrr(r6r!r rs$ 65-}/f}/@#EE4 O$9=C IICehhoo.5CYY C%CL%bii%D%r!