o ei$'@sUdZddlZddlZddlmZddlmZddlZdae e d<de de fdd Z dee fd d Z      dd dZdefddZdefddZGdddZddZddZd ddZddZddZdS)!zGuard for running certain operations on main process only Authors: * Abdel Heba 2020 * Aku Rouhe 2020 * Peter Plantinga 2023 * Adel Moumen 2024 Nwraps)OptionalMAIN_PROC_ONLYmessagereturncCs"t}|durd|d|S|S)zPrefix a message with the rank of the process. Arguments --------- message : str The message to prefix. Returns ------- str The message prefixed with the rank, if known. Nz[rank: z] )get_rank)rrankr [/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/speechbrain/utils/distributed.pyrank_prefixed_messages r cCs2d}|D]}tj|}|durt|SqdS)a|Get the rank of the current process. This code is taken from the Pytorch Lightning library: https://github.com/Lightning-AI/pytorch-lightning/blob/bc3c9c536dc88bfa9a46f63fbce22b382a86a9cb/src/lightning/fabric/utilities/rank_zero.py#L39-L48 Returns ------- int or None The rank of the current process, or None if the rank could not be determined. )RANK LOCAL_RANK SLURM_PROCIDJSM_NAMESPACE_RANKN)osenvirongetint) rank_keyskeyr r r r r's   rFcCs|durg}|dur i}|durg}|duri}t||i|t|durB|r3||i|dSts=||i|tdSdS)aRuns a function with DPP (multi-gpu) support. The main function is only run on the main process. A post_function can be specified, to be on non-main processes after the main func completes. This way whatever the main func produces can be loaded on the other processes. Arguments --------- func : callable Function to run on the main process. args : list, None Positional args to pass to func. kwargs : dict, None Keyword args to pass to func. post_func : callable, None Function to run after func has finished on main. By default only run on non-main processes. post_args : list, None Positional args to pass to post_func. post_kwargs : dict, None Keyword args to pass to post_func. run_post_on_main : bool Whether to run post_func on main process as well. (default: False) N)main_process_only ddp_barrierif_main_process)funcargskwargs post_func post_args post_kwargsrun_post_on_mainr r r run_on_main=s"# r!cCstjo tjS)z2Returns whether the current system is distributed.)torch distributed is_availableis_initializedr r r r is_distributed_initializedwsr&cCstr tjdkSdS)z8Returns whether the current process is the main process.rT)r&r"r#rr r r r rsrc@s eZdZdZddZddZdS)MainProcessContextz Context manager to ensure code runs only on the main process. This is useful to make sure that `MAIN_PROC_ONLY` global variable is decreased even if there's an exception raised inside of `main_proc_wrapped_func` fn. cCs td7a|S)z(Enter the context. Increase the counter.r)selfr r r __enter__szMainProcessContext.__enter__cCs td8adS)z'Exit the context. Decrease the counter.r(Nr))r*exc_type exc_value tracebackr r r __exit__s zMainProcessContext.__exit__N)__name__ __module__ __qualname____doc__r+r/r r r r r's r'cstfdd}|S)zFunction decorator to ensure the function runs only on the main process. This is useful for things like saving to the filesystem or logging to a web address where you only want it to happen on a single process. csPttr|i|WdS WddS1s!wYdS)z>This decorated function runs only if this is the main process.N)r'r)rrfunctionr r main_proc_wrapped_funcs  $z1main_process_only..main_proc_wrapped_funcr)r5r6r r4r rsrcCsNtdksts dStjtjjjkr tjjtj gddStjdS)a7 Synchronize all processes in distributed data parallel (DDP) mode. This function blocks the execution of the current process until all processes in the distributed group have reached the same point. It ensures that no process moves ahead until every other process has also reached this barrier. If DDP is not being used (i.e., only one process is running), this function has no effect and immediately returns. Returns ------- None Example ------- >>> ddp_barrier() >>> print("hello world") hello world r(N) device_ids) rr&r"r# get_backendBackendNCCLbarriercudacurrent_devicer r r r rs rcCs0tdksts |S|g}tjj||d|dS)aIn DDP mode, this function will broadcast an object to all processes. Arguments --------- communication_object: Any The object to be communicated to all processes. Must be picklable. See docs for ``torch.distributed.broadcast_object_list()`` src: int The rank which holds the object to be communicated. Returns ------- The communication_object passed on rank src. r()srcr)rr&r"r#broadcast_object_list)communication_objectr>communication_listr r r ddp_broadcasts rBcCs&tdksts |Stjj||d|S)a`In DDP mode, this function will perform an all_reduce operation with the specified torch operator. See: https://pytorch.org/docs/stable/distributed.html#torch.distributed.all_reduce Arguments --------- communication_object: Any The object to be reduced across processes. reduce_op: torch.distributed.ReduceOp The operation to perform. E.g. include torch.distributed.ReduceOp.AVG or torch.distributed.ReduceOp.SUM. See the Torch documentation for more. Returns ------- The communication_object once reduced (or itself if DDP not initialised) r()op)rr&r"r# all_reduce)r@ reduce_opr r r ddp_all_reducesrFcCs&tjd}tjd}|dus|durdSt|}|ddks2|dtjkr2tdtdt|}|dd krFtj sEtd n(|ddkrVtj sUtd n|dd krftj setd nt|dd|dd krt d|}tj|tj j|d|tjddddS)aThis function will initialize the ddp group if distributed_launch bool is given in the python command line. The ddp group will use distributed_backend arg for setting the DDP communication protocol. `RANK` Unix variable will be used for registering the subprocess to the ddp group. Arguments --------- run_opts: list A list of arguments to parse, most often from `sys.argv[1:]`. Returns ------- None r rNdistributed_backendgloor(zKilling process z Not enough GPUs available!ncclz&NCCL is not supported in your machine.z&GLOO is not supported in your machine.mpiz%MPI is not supported in your machine.z& communication protocol doesn't exist.zcuda:i )seconds)backendr timeout)rrrrr"r< device_count ValueErrorstrr#is_nccl_availableis_gloo_availableis_mpi_availabledevice set_deviceinit_process_groupdatetime timedelta)run_optsr local_rankrTr r r ddp_init_groupsH               r[)NNNNNF)r)r3rWr functoolsrtypingrr"rr__annotations__rPr rr!boolr&rr'rrrBrFr[r r r r s0     :