Vh5FddlZddlZddlZddlmZddlmZddlmZm Z m Z ddl m Z ddl Z ddlmZddlmZmZddlmZddlmZdd lmZdd lmZdd lmZdd lmZdd l m!Z!m"Z"ddl#m$Z$m%Z%ddl&m'Z'ddl(m)Z)ddl*m+Z+ddl,m-Z-ddl.m/Z/m0Z0m1Z1gdZ2GddeZ3e de4 d-de"de+de ejjde6de7d e e%d!e!fd"Z8ed#$e/dddddd%de"d&e e9ejtdfde e+d e e%de ejjde7d!e!fd'Z;ed#$dddde3jxd(de"d&e e9ejtdfde e+d e e%de ejjd)e3d!efd*Z=de"d!e"fd+Z> d-de"de+de ejjde6de7d e e%d!e!fd,Z?y).N)Future)Enum)castOptionalUnion) deprecated)_copy_state_dict_create_cpu_state_dict)_AsyncCheckpointExecutor)$_ProcessBasedAsyncCheckpointExecutor)#_ThreadBasedAsyncCheckpointExecutor)_storage_setup)DefaultSavePlanner)_dcp_method_logger)MetadataSTATE_DICT_TYPE)SavePlan SavePlanner) AsyncStager)Stateful) StorageWriter)_get_default_group) _api_bc_check _DistWrapper_profile)save_state_dictsave async_saveAsyncCheckpointerTypeceZdZdZdZdZy)r z!Enum for async checkpointer type.threadprocessN)__name__ __module__ __qualname____doc__THREADPROCESS]/home/dcms/DCMS/lib/python3.12/site-packages/torch/distributed/checkpoint/state_dict_saver.pyr r 's+ FGr+r za`save_state_dict` is deprecated and will be removed in future versions.Please use `save` instead.)categoryF state_dictstorage_writer process_groupcoordinator_rankno_distplannerreturnc |jt5t||||||cdddS#1swYyxYw)z3This method is deprecated. Please switch to 'save'.N)resetr_save_state_dict)r.r/r0r1r2r3s r,rr.sF             s5>T)log_exceptions) checkpoint_idr/r3r0r2r9c xtjjd|xs,tj xstj }|rt jdt5ttt||d}tt|||||cdddS#1swYyxYw)a Save a distributed model in SPMD style. This function is different from ``torch.save()`` as it handles ``ShardedTensor`` , and ``DTensor`` by having each rank only save their local shards. For each ``Stateful`` object (having both a ``state_dict`` and a ``load_state_dict``), save will call ``state_dict`` before serialization. .. warning:: There is no guarantees of Backwards Compatibility across PyTorch versions for saved state_dicts. .. warning:: If using the `process_group` argument, make sure that only its ranks call `save_state_dict` and that all data in state_dict belong to it. .. note:: When saving checkpoint for FSDP's `ShardingStrategy.HYBRID_SHARD`, only one of the shard_group should be calling `save_state_dict` and the corresponding process group needs to be passed in. .. note:: If no process group is available, this function assumes the intention is to save the state_dict in the local process. .. note: Rank 0 is assumed to be the coordinator rank. Args: state_dict (Dict[str, Any]): The state_dict to save. checkpoint_id (Union[str, os.PathLike, None]): The ID of this checkpoint instance. The meaning of the checkpoint_id depends on the storage. It can be a path to a folder or to a file. It can also be a key if the storage is a key-value store. (Default: ``None``) storage_writer (Optional[StorageWriter]): Instance of StorageWriter used to perform writes. If this is not specified, DCP will automatically infer the writer based on the checkpoint_id. If checkpoint_id is also None, an exception will be raised. (Default: ``None``) planner (Optional[SavePlanner]): Instance of SavePlanner. If this is not specificed, the default planner will be used. (Default: ``None``) process_group (Optional[ProcessGroup]): ProcessGroup to be used for cross-rank synchronization. (Default: ``None``) no_dist (bool): If ``True``, this function will assume the intent is to load a checkpoint without using cross-rank synchronization. (Default: ``False``) Returns: Metadata: Metadata object for the saved checkpoint. Example: >>> # xdoctest: +SKIP >>> my_model = MyModule() >>> state_dict = {"model": my_model} >>> fs_storage_writer = torch.distributed.checkpoint.FileSystemWriter( ... "/checkpoint/1" ... ) >>> torch.distributed.checkpoint.save( >>> state_dict=state_dict, >>> storage_writer=fs_storage_writer, >>> ) .. note:: save_state_dict uses collectives to coordinate writes across ranks. For NCCL-based process groups, internal tensor representations of objects must be moved to the GPU device before communication takes place. In this case, the device used is given by ``torch.cuda.current_device()`` and it is the user's responsibility to ensure that this is set so that each rank has an individual GPU, via ``torch.cuda.set_device()``. z!torch.distributed.checkpoint.savezptorch.distributed is disabled, unavailable or uninitialized, assuming the intent is to save in a single process.Freader)r.r/r0r2r3N)torch_C_log_api_usage_oncedist is_availableis_initializedwarningswarnrrrrr7_stateful_to_state_dict)r.r9r/r3r0r2s r,rrJsr HH  !DEQd//11Q4;N;N;P7PG ~     >.-PUV  .z:)'     s 15B00B9)r9r/r3r0async_checkpointer_typerFctjjdtjrJtj r6|xs t }tjd|jvsJdttt||d}t|}t|tr|j|}nt!|}t#||d|t$j&k(r t)n t+}|j-|||||} t|tr|j.r|j1| S)aAsynchronous version of ``save``. This code first de-stages the state_dict on to the staging storage (defaults to CPU memory), and then calls the `save` in a separate thread. .. warning:: This feature is experimental and subject to change. Args: state_dict (Dict[str, Any]): The state_dict to save. checkpoint_id (Union[str, os.PathLike, None]): The ID of this checkpoint instance. The meaning of the checkpoint_id depends on the storage. It can be a path to a folder or to a file. It can also be a key if the storage is a key-value store. (Default: ``None``) storage_writer (Optional[StorageWriter]): Instance of StorageWriter used to perform 'stage' and 'save'. If this is not specified, DCP will automatically infer the writer based on the checkpoint_id. If checkpoint_id is also None, an exception will be raised. (Default: ``None``) planner (Optional[SavePlanner]): Instance of SavePlanner. If this is not specificed, the default planner will be used. (Default: ``None``) process_group (Optional[ProcessGroup]): ProcessGroup to be used for cross-rank synchronization. (Default: ``None``) Returns: Future: A future holding the resultant Metadata object from `save`. Example: >>> # xdoctest: +SKIP >>> my_model = MyModule() >>> state_dict = {"model": my_model} >>> fs_storage_writer = torch.distributed.checkpoint.FileSystemWriter( ... "/checkpoint/1" ... ) >>> checkpoint_future = torch.distributed.checkpoint.async_save( >>> state_dict=state_dict, >>> storage_writer=fs_storage_writer, >>> ) >>> >>> # ... do some work ... >>> >>> checkpoint_future.result() z'torch.distributed.checkpoint.async_savecpuzfA CPU backend must be enabled for async save; try initializing process group with 'cpu:gloo,cuda:nccl'Fr;) type_check)r9r/r3r0)r=r>r?r@rArBrdevice _device_typesrrrrE isinstancerstager r r r)r r execute_save should_synchronize_after_executesynchronize_staging) r.r9r/r3r0rFpgstaged_state_dictexecutorfs r,rrs0r HH  !JK t224  202 LL 2#3#3 3  u 3 ~nmERN)4J.+.*00<2:>%65I #&;&C&C C -. 0 2 %%#%# &A >;/  ; ;**, Hr+ci}|jD]*\}}t|tr|jn|||<,|S)z]Creates a shallow copy of `state_dict` where `state_dict` is called for each Stateful object.)itemsrLrr.)r.stateful_state_dictkeyelems r,rErEsK%%' T!+D(!;DOO  C   r+c tjjdt|| |  t Jdi}t ddx}||d< j |d<td i| fd}td i|fd} jd||  td i| fd} td i|fd} jd | | S) Nz,torch.distributed.checkpoint.save_state_dictr9r0cJj}dtjjjvr2t j djjnj|jjjj}j|}|S)N storage_metazThe function definition for SavePlanner.set_up_planner has been updated to include the storage_meta argument. Please update your implementation to include this parameter.)r.r\is_coordinator) r\inspect signatureset_up_planner parametersrCrDr]set_up_storage_writercreate_local_planprepare_local_plan)r\ local_plandistWr3r.r/s r, local_stepz$_save_state_dict..local_step>s"""%224 !2!273I3I!J!U!U U MM.   " ":u/C/C D  " "%)$33 #  ,,U-A-AB..0 #66zB r+cZJj|\}j|}|SN)create_global_planprepare_global_plan)all_local_plansglobal_metadatar3r/s r, global_stepz%_save_state_dict..global_stepUs<"""+2+E+Eo+V((<<_Mr+plancJj}j|}|j|jSri) finish_plan write_datawaitvalue)final_local_plan all_writes central_planr3r/s r,rrz$_save_state_dict..write_data`sL""""..|<#../?I !!r+c6Jj|S)N)metadataresults)finish) all_resultsrmr/s r,finish_checkpointz+_save_state_dict..finish_checkpointis(*** Lr+writer*) r=r>r?rrgetattrgrouprreduce_scatter all_reduce)r.r/r0r1r2r3 ckpt_kwargsckpt_idrgrnrrr}rwrfrms`` ` @@@r,r7r7(s HH  !OP G 5E FE$&   OK>?DAAN'. O$',{{ O$&+&',&+&'#11&*kRL&+&"'"&+&'   GZ1B CCr+)NrFN)@r^osrCconcurrent.futuresrenumrtypingrrrtyping_extensionsrr=torch.distributed distributedr@#torch.distributed._state_dict_utilsr r ,torch.distributed.checkpoint._async_executorr 4torch.distributed.checkpoint._async_process_executorr 3torch.distributed.checkpoint._async_thread_executorr +torch.distributed.checkpoint._storage_utilsr,torch.distributed.checkpoint.default_plannerr#torch.distributed.checkpoint.loggerr%torch.distributed.checkpoint.metadatarr$torch.distributed.checkpoint.plannerrr$torch.distributed.checkpoint.stagingr%torch.distributed.checkpoint.statefulr$torch.distributed.checkpoint.storager"torch.distributed.distributed_c10drutilsrrr__all__r FutureWarning ProcessGroupintboolrstrPathLikerr(rrEr7r*r+r,rs %((( XGKBKF<:>A88 MD ! 26%)   ! D--.     k "     .4(48.2%)15j j bkk4/0j ]+ j k " j D--. j j j )j Z4(48.2%)155J5Q5Qa a bkk4/0a ]+ a k " a D--. a 3a  a )a HO26%) GDGD!GDD--.GD GD  GD k " GDGDr+