Mjn UddlZddlZddlZddlZddlZddlmZmZddlm Z ddl m Z m Z ddl Z ddlmZddlmZddlmZmZddlmZdd lmZdd lmZmZmZmZmZmZm Z m!Z!dd l"m#Z#m$Z$m%Z%m&Z&m'Z'm(Z(m)Z)dd l*m+Z+m,Z,m-Z-m.Z.m/Z/m0Z0m1Z1dd l2m3Z3ddl4m5Z5ddl6m7Z7ejpe9Z:ejveGdde$Z?Gdde?Z@ d-deAeBe fdedeCde#fdZDdeEe#deEe#fdZFdeAeBe fd eCde&fd!ZG d-deEe&d"eCdeHeEe&effd#ZIdedefd$ZJd%ed&edeCfd'ZKd(e jd)edeCfd*ZMd+eEe&dedeEeBfd,ZNy).N) bisect_rightinsort)ChainMap)Anycast)narrow_tensor_by_index)dedup_save_plans)FLATTEN_MAPPINGflatten_state_dict)_flatten_sharded_tensors) set_element)BytesStorageMetadataChunkStorageMetadataMetadata MetadataIndexSTATE_DICT_TYPE STORAGE_TYPES StorageMetaTensorStorageMetadata)LoadPlan LoadPlannerReadItemSavePlan SavePlanner WriteItem WriteItemType)_compare_save_plans_contains_usable_plan"_create_default_metadata_only_plan_create_read_items_create_write_items_init_state_dict_merge_delta_local_plans)find_state_dict_object)DTensor)_versionlogger)DefaultSavePlannerDefaultLoadPlannercreate_default_local_load_plancreate_default_global_load_plancreate_default_local_save_plancreate_default_global_save_planc eZdZUeed< ddedededzdededdf d Z dd ed edzd eddfd Z de fdZ de e de e fdZ de e dee e effdZde e dee e e e effdZde e dee e effdZde de fdZde de fdZdedej.ej2zfdZdedefdZdedefdZy)r)mappingsNr flatten_sharded_tensorsdedup_replicated_tensorsdedup_save_to_lowest_rankenable_plan_cachingreturnc||_||_i|_||_|tj d|j j|_||_ y)NzDefaultSavePlanner's `dedup_replicated_tensors` argument is being deprecated, and no longer has any effect. Please remove this argument from your call.) r r1r0r3r(warning __class____name___cached_plans_key_enable_plan_caching)selfr r1r2r3r4s k/var/www/html/ia-enhanced/venv/lib/python3.12/site-packages/torch/distributed/checkpoint/default_planner.py__init__zDefaultSavePlanner.__init__IsV#5'>$ )B& # / NN"  '+nn&=&=$7! state_dict storage_metais_coordinatorc|jrt|\}|_|jr t|}||_||_yN)r r0r1r r@rB)r<r@rArBs r=set_up_plannerz!DefaultSavePlanner.set_up_planner^s@  " "(::(F %J  ' '1*=J$,r?ct|j|j}|jr!t j ||j }||_|jrl|jtjvrIt|tj|jr"tjdtgdS||_|jS)N planner_datazINo change in the local plan. Skipping sending the plan to the coordinatorFusable)r-r@rBr dataclassesreplacer0planr;r:r_cached_save_planrr(infor_pending_local_plan)r<rMs r=create_local_planz$DefaultSavePlanner.create_local_planks-doot?R?RS  " "&&t$--HD  $ $&&+*G*GG'+778N8NO _ 511,0(yyr? all_plansc.t||jSrD)r r3)r<rRs r=_dedup_save_plansz$DefaultSavePlanner._dedup_save_planss 4+I+IJJr?cr|j|}t|\}}|jrA|Dcgc]}|j}}t t |}t j||}t||}|r5dj|} t| dkDr| dddz} td| ||fScc}w)NrGz; iz... (truncated)z Failed to validate global plan: ) rTr.r rHdictrrKrL_validate_global_planjoinlen ValueError) r<rR deduped_plans global_planmetadatapplanner_data_dictmerged_mappingsvalidation_errors error_summarys r=_create_global_planz&DefaultSavePlanner._create_global_plans..y9 ? N X  " " :E EA E  E"8->#?@O"**8/RH1+xH  II&78M=!C' -ds 36G G ? OP PH$$!FsB4cg}|jtjvrp|j|\}}|tj|j<|tj|j<|tj |j<|||fSt |sVtgdgt|z}tj|j}tj |j}nttj|j|}|tj|j<|j|\}}|j|jvrittj|j|D]?\}}t||r|jtgd/|j|A|tj|j<|tj |j<|||fS)zw Create global plan with caching. Returns a tuple of global_plan_delta, global_plan, metadata. FrI) r:r_cached_all_plansrc_cached_global_plan_cached_metadatarrrYr#ziprappend)r<rRglobal_plan_deltar\r] merged_plans cached_plannew_plans r= _create_global_plan_with_cachingz3DefaultSavePlanner._create_global_plan_with_cachings-/  ! !)F)F F %)$<$rrrErrQlistrTtuplerrcrnrprurxrrTensorrrrrrr{r}r?r=r)r)Fs$((,04*/$) 8 8"&8#'+ 8 $( 8 " 8 80,0$ -# -"D( - -  -88K4>Kd8nK%h% tH~x' (%2:8h:8 tH~tH~x7 8:8x0h0 tH~x' (0,(xH&9y9U\\BJJ5N9>=>S>9cr?r)c FeZdZUdZeed<eed< ddedededdfd Z dd ed e dzd eddfd Z de fdZ de e de e fdZde de fdZdedej$ddfdZdefdZdedej,ddfdZdedej,fdZdedej,fdZy)r*ak DefaultLoadPlanner that adds multiple features on top of LoadPlanner. In particular it adds the following: flatten_state_dict: Handle state_dict with nested dicts flatten_sharded_tensors: For FSDP in 2D parallel mode allow_partial_load: If False, will raise a runtime error if a key is present in state_dict, but not in the checkpoint. original_state_dictr0r r1allow_partial_loadr5NcJ||_||_i|_i|_||_yrD)r r1rr0r)r<r r1rs r=r>zDefaultLoadPlanner.__init__1s+ #5'>$#%  "4r?r@r]rBct|||_|jr t|}|jrt |\}|_||_||_||_yrD) r"rr1r r r0r@r]rB)r<r@r]rBs r=rEz!DefaultLoadPlanner.set_up_planner=sW $#-  ' '1*=J  " "(::(F %J $  ,r?c|j td|jrt|jj }t|jj j }||z }|r[dt_t|j\}}t|j }||zr||c|_|_ dt_t|j|j|j S)Nzself.metadata is not None2_3) r]AssertionErrorr setr@keysstate_dict_metadatar'_derived_versionrr0r+r)r< current_keys load_keys missing_keysold_state_dict old_mappingsold_keyss r=rQz$DefaultLoadPlanner.create_local_planPs ==  !<= =  " "$t3356LDMM==BBDEI$|3L,1)/A,,0, ~2245l*5C\2DOT]-1)- OOT]]0G0G,G  r?r\ct|SrD)r,)r<r\s r=rpz%DefaultLoadPlanner.create_global_planxs .{;;r?rmc|SrDr)r<rms r=rxzDefaultLoadPlanner.finish_plan{sr? read_itemvaluec(|jrNt|j|j|jj t j|dyt j|d|j|jj <y)NF) weights_only) r r rr0 dest_indexfqnrloadr@)r<rrs r= load_byteszDefaultLoadPlanner.load_bytes~sl  " " (( i22667 5u5  9> E9DOOI0044 5r?c\|j|j}|j||SrD) lookup_tensorrtransform_tensorr<rtensors r=resolve_tensorz!DefaultLoadPlanner.resolve_tensors+##I$8$89$$Y77r?rcyrDrrs r= commit_tensorz DefaultLoadPlanner.commit_tensors r?r|c.t|j|Srrrs r=rz DefaultLoadPlanner.lookup_tensorrr?cDt||j|jSr)r dest_offsetslengthsrs r=rz#DefaultLoadPlanner.transform_tensors%fi.D.DiFWFWXXr?)TTFr)r9rr__doc__rrr rr>rrErrQrrprxrrrrrrrrrrrrr?r=r*r*#s6)($((,#( 5  5"& 5! 5  5%)$ -#-T/- -  -&& 8& P=>U\\>Y(YELLYr?r*c `eZdZdZd fd ZdededefdZ d de dedzd eddffd Z xZ S) _EmptyStateDictLoadPlannera Extension of DefaultLoadPlanner, which rebuilds state_dict from the saved metadata. Useful for loading in state_dict without first initializing a model, such as when converting a DCP checkpoint into a Torch save file. . N.B. `state_dict` must be an empty dictionary when used with this LoadPlanner .. warning:: Because the entire state dict is initialized, It's recommended to only utilize this LoadPlanner on a single rank or process to avoid OOM. Nc2||_t||i|yrD)rsuperr>)r<rargskwargsr8s r=r>z#_EmptyStateDictLoadPlanner.__init__s  $)&)r?keyr]r5c 4jy|jvryg}|jj|}|D]D}|r/|jdj |dt |g4|j|Ft fd|Dryy)NT.c3:K|]}|jvywrD)r).0unflattened_keyr<s r= zA_EmptyStateDictLoadPlanner._should_include_key..sT$))+TsF)rrHgetrirXstrany)r<rr]unflattened_keysrHrs` r=_should_include_keyz._EmptyStateDictLoadPlanner._should_include_keys 99  $)) &(,,005 + 9O ''HH.r2C4HIJ !''8 9 TCST Tr?r@rBc|r td| td|jjD]\}}|j||st |t r5t j|j|jj}|j)||jvrt||j|||||<t|9|||y)Nznot state_dictzmetadata is not None)dtype)rritemsr isinstancerremptysize propertiesrrHr rrE)r<r@r]rBkvr8s r=rEz)_EmptyStateDictLoadPlanner.set_up_planners  !12 2   !78 800668 "DAq++Ax8!23KKall.@.@A$$0Q(:O:O5OJ(=(=a(@!D ! 1  " z8^Dr?rDr) r9rrrr>rrrrrrE __classcell__)r8s@r=rrsi *sh44%)$ E#ET/E E  EEr?rr@r]strictr5c g} |jD]\}}||jvr|rtd|d&|j|}t|trVt |ddI|j |j k7r,td|j d|j d|t|tr,|jj|t|||z }|t|||z }t|S)Nz&Missing key in checkpoint state_dict: rrzSize mismatch between saved z and current: z for ) rr RuntimeErrorrrgetattrrrZr% device_meshget_coordinater r)r@r]rrequestsrobjmds r=r+r+sH$$&9S h22 2"%KC5PQ#RSS  ) )# . r0 1VT*6388:%.rwwi~chhj\QVWZV[\  c7 #--/;.sB<< *3C8 8H/92 H r?rRc|S)z Create global load plan used by DefaultLoadPlanner. The default load behavior involved no global coordination and this function currently doesn't change the local plans. r)rRs r=r,r,s  r?rBcg}|jD]O\}}t|tr+|jj 1|t ||z }A|t ||z }Qt |S)a Create the ``SavePlan`` used by DefaultSavePlanner. On non-coordinator ranks, this function ignores tensors and non-tensor objects, only producing writes for ShardedTensor objects. On the coordinator rank, produce writes for all values. )rrr%rrr!r)r@rBrrrs r=r-r-swH$$& 6S c7 #--/;/S99 +C5 5H 6 H r?rewrite_index_hintsc Zi}g}|D]}g}|jD]}|jtjk7r#|jj |vr t d|jtjk(r3t||jj <|j||j t dtt|j|jj t|jj|jjg}|}|rKt!j"|jt%|j&} t!j"|| }|j||jj(#t d|jj d|j&j|jj(|jt!j"|||t+|fS)a6 Create the global plan and metadata used by DefaultSavePlanner. Metadata is produced by concatenating the metadata of all ``WriteItem`` from the supplied plans. The only global planning change is to update index hints in all ``MetadataIndex`` objects if ``rewrite_index_hints`` is True. zitem.index.fqn not in mdzitem.tensor_data is not None)rrchunks)r|zZ Cannot create MD for tensor without bounds. FQN: z )r)rrrSHARDr|rrrrri tensor_datarr setdefaultrrrKrLrYrchunkr) rRrr new_plansrM new_itemsitem tensor_mdnew_item new_indexs r=r.r..s$&BI&E JJ# @DyyM///::>>R'()CDDyyM111%9%;4::>>"  &##+()GHH )MM -'+'7'7'B'B!%!1!1!6!6#%   & + 3 3 #i.>.>*?!I +224yIH  *##))1(.**..)**  ''(8(8(>(>?G# @H ,,TCDM&EN x| $$r?c:t|}t|g\}}|S)zTReturn the ``Metadata`` if DefaultSavePlanner was used to checkpoint ``state_dict``.)rr.)r@rM_rs r=_create_default_local_metadatarfs! -j 9D +TF 3EAr Ir?box0box1ct|j}t|D]d}|j||j||j|zk\ry|j||j||j|zk\sdyy)z9Check if two boxes overlap. Tuples are (offset, lengths).FT)rYoffsetsrangesizes)rrndimsis r=_check_box_overlaprmsz  E 5\ <<?dll1o 1 = = <<?dll1o 1 = =  r?outer_box_size inner_boxctt|D]Q}|j|dkry|j|dkry|j||j|z||kDsQyy)NrFT)rrYrr)rrrs r=_check_box_boundsr}su3~& '   Q ! # ??1  !   Q )//!"4 4~a7H H  r?r\c g}|jjD].\}ttrt j dk(r1j d}D]u}tj |s;d|dj d|}tj||j||tj|jz }wt dkDrt j ttdfdt!tt fd }g}|D]} | } | j"} | | jz} t%|| t&j(f} | r|d | =|D]H\}}|}t+| |sd|d | d |}tj||j|Jt-|| | ftjj }t |dkDs||k7sd|d |d|}tj||j|1|S)zNValidate the global plan and return a list of error messages (empty if valid).rzkey:z& has out of bounds chunk: tensor-size:z chunk: r&c"j|SrD)r)drs r=z'_validate_global_plan..s%**Q-r?)defaultrcXjgfdtDS)Nc3BK|]}j|ywrD)r)rrridxs r=rz:_validate_global_plan....s BfSk))!,Bs)rr)r rdims sweep_dims`r=rz'_validate_global_plan..s-3K'' 2!BeDkB!r?)rNz has overlapping chunks:  z invalid fill tensor-volume: z chunks-volume: )rrrrrYrrrr(r7rimathprodrmaxrsortedrrsysmaxsizerr)r\r]errorsr chunks_volumermsgsorted_indicesactiver currentstartendcutoffr other_idxother tensor_volumerr r rs @@@@r=rWrWsKF2288:7 U e1 2  uzz?a    4E$UZZ73% ##(::,hug?s# c" TYYu{{3 3M 4 v;?uzz?DE$K8OPI#c&k"N-/F% + + 2gmmI66%fuckk.BCww$*+LAy"9-E)'59 $SE)B7)1UGTs+ c* +vSz*! +& %**- { a M]$Bse8 /!1-B  NN3  MM# o7r Mr?)T)OrKrloggingrrbisectrr collectionsrtypingrrrtorch.distributed._shard._utilsr.torch.distributed.checkpoint._dedup_save_plansr )torch.distributed.checkpoint._nested_dictr r 2torch.distributed.checkpoint._sharded_tensor_utilsr &torch.distributed.checkpoint._traverser %torch.distributed.checkpoint.metadatarrrrrrrr$torch.distributed.checkpoint.plannerrrrrrrr,torch.distributed.checkpoint.planner_helpersrrrr r!r"r#"torch.distributed.checkpoint.utilsr$torch.distributed.tensorr%r' getLoggerr9r(Loggerr__all__r)r*rrVrrr+rr,r-rr.rrSizerrWrr?r=r4s  '  BKX>   F,+**844 ZZztYtYn@E!3@EHDH'S#X'*2'<@' 'T H~  (^ S#X04 :!%5%H~5%5% 4>8 #$5%p8 1 9M RV  JJ +?   <tH~<<dSVi<r?