a dp@sBddlZddlZddlZddlZddlZddlmZddlmZddl m Z m Z m Z m Z mZmZmZmZmZmZddlZddlmZddlmZddlmZddlmZmZdd lmZdd l m!Z!dd l"m#Z#dd l$m%Z%dd l&m'Z'ddl(m)Z)m*Z*ddl+m,Z,ddl-m.Z.edZ/er,e/r,ddl0Z0Gddde#e%Z1dS)N)contextmanager)Path) AnyDict GeneratorIterableListMappingOptionalTuple TYPE_CHECKINGUnion)RequirementCache)Module) Optimizer) AcceleratorCUDAAccelerator)ClusterEnvironment) Precision) DDPStrategy)_Sharded)log)rank_zero_inforank_zero_only) reset_seed)_PATH deepspeedc/s&eZdZdZdddddddddd d dd d d ddddddddddddddejddddddddddddddf,eeee e eee e e e e e e e e eee ee eeeee e eee ee e e e effe eeejeeee e e e eeeeeeeee dd-fdd ZeedddZee e e fdddZeddddZeeeedeefdd d!Zedd"d#d$Zeed%d&d'Ze e!d(dd)d*Z"dRe e eedd+d,d-Z#e e e efd.d/d0Z$e ee%efe&e efdd1d2d3Z'ee&e efdd4d5d6Z(e)e dd7d8d9Z*dSeeeedefd:d;d<Z+ddd=d>Z,ddd?d@Z-dddAdBZ.dddCdDZ/dddEdFZ0dddGdHZ1eeee eeeeeee e e e e e e ee e eee ee dIdJdKZ2ee&e efddLdMdNZ3ee e e e effee e efdOdPdQZ4Z5S)TDeepSpeedStrategyZPL_DEEPSPEED_CONFIG_PATHNTcpuFz /local_nvmeiiʚ;ilJ)i ri)- acceleratorzero_optimizationstage remote_deviceoffload_optimizeroffload_parametersoffload_params_device nvme_pathparams_buffer_countparams_buffer_size max_in_cpuoffload_optimizer_deviceoptimizer_buffer_count block_size queue_depth single_submitoverlap_events thread_count pin_memorysub_group_sizecontiguous_gradients overlap_commallgather_partitionsreduce_scatterallgather_bucket_sizereduce_bucket_sizezero_allow_untested_optimizerlogging_batch_size_per_gpuconfig logging_levelparallel_devicescluster_environment loss_scaleinitial_scale_powerloss_scale_window hysteresismin_loss_scalepartition_activationscpu_checkpointingcontiguous_memory_optimizationsynchronize_checkpoint_boundaryload_full_weights precisionprocess_group_backendreturnc-. sts tdtj||| |+|,dd|_|||_|jdur|j|||||||| | | || | ||||||&|'|(|)||||||||d|_ddl}-d|_ |-j j j |||_|*|_|!|_|"|_|#|_|$|_|%|_d|_dS)aProvides capabilities to run training using the DeepSpeed library, with training optimizations for large billion parameter models. `For more information: https://pytorch- lightning.readthedocs.io/en/stable/advanced/model_parallel.html#deepspeed`. .. warning:: ``DeepSpeedStrategy`` is in beta and subject to change. Defaults have been set to enable ZeRO-Offload and some have been taken from the link below. These defaults have been set generally, but may require tuning for optimum performance based on your model size. `For more information: https://www.deepspeed.ai/docs/config-json/#zero-optimizations-for-fp16-training`. Arguments: zero_optimization: Enable ZeRO optimization. This is compatible with either ``precision=16`` or ``precision="bf16"``. stage: Different stages of the ZeRO Optimizer. 0 is disabled, 1 is optimizer state partitioning, 2 is optimizer+gradient state partitioning, 3 is optimizer+gradient_parameter partitioning using the infinity engine. remote_device: Device to instantiate the model on initially (``cpu`` or ``nvme``). offload_optimizer: Enable offloading optimizer memory and computation to CPU or NVMe based on ``offload_optimizer_device``. offload_parameters: When using ZeRO Stage 3, Enable offloading parameter memory and computation to CPU or NVMe based on ``offload_params_device``. offload_params_device: When offloading parameters choose the device to offload to, ``cpu`` or ``nvme``. offload_optimizer_device: When offloading optimizer state choose the device to offload to, ``cpu`` or ``nvme``. params_buffer_count: Number of buffers in buffer pool for parameter offloading when ``offload_params_device`` is ``nvme``. params_buffer_size: Size of buffers in buffer pool for parameter offloading when ``offload_params_device`` is ``nvme``. max_in_cpu: Number of parameter elements to maintain in CPU memory when offloading to NVMe is enabled. nvme_path: Filesystem path for NVMe device for optimizer/parameter state offloading. optimizer_buffer_count: Number of buffers in buffer pool for optimizer state offloading when ``offload_optimizer_device`` is set to to ``nvme``. This should be at least the number of states maintained per parameter by the optimizer. For example, Adam optimizer has 4 states (parameter, gradient, momentum, and variance). block_size: When using NVMe Offloading, the I/O block size in bytes. queue_depth: When using NVMe Offloading, the I/O queue depth. single_submit: When using NVMe Offloading, submit requests to storage device as multiple individual requests, as opposed to one block of requests. overlap_events: When using NVMe Offloading, submit requests to storage device in an overlapped fashion without waiting for completion of earlier requests. thread_count: When using NVMe Offloading, Intra-request parallelism for each read/write submitted by a user thread. pin_memory: When using ZeRO stage 3, pin optimizer state memory on CPU. This could boost throughput at the cost of extra memory overhead. sub_group_size: When using ZeRO stage 3, defines the number of parameters within a sub group to offload at a time. Smaller numbers require more communication, but improve memory efficiency. contiguous_gradients: Copies gradients to a continuous buffer as they are produced. Avoids memory fragmentation during backwards. Useful when training large models. overlap_comm: Overlap the reduction (synchronization) of gradients with the backwards computation. This is a speed optimization when training across multiple GPUs/machines. allgather_partitions: All gather updated parameters at the end of training step, instead of using a series of broadcast collectives. reduce_scatter: Use reduce/scatter instead of allreduce to average gradients. allgather_bucket_size: Number of elements to allgather at once. Used to limit the memory required for larger model sizes, with a tradeoff with speed. reduce_bucket_size: Number of elements to reduce at once. Used to limit the memory required for larger model sizes, with a tradeoff with speed. zero_allow_untested_optimizer: Allow untested optimizers to be used with ZeRO. Currently only Adam is a DeepSpeed supported optimizer when using ZeRO. logging_batch_size_per_gpu: Config used in DeepSpeed to calculate verbose timing for logging on a per sample per second basis (only displayed if logging=logging.INFO). To obtain accurate logs when using datasets that do not support batch samplers, set this to the actual per gpu batch size. config: Pass in a deepspeed formatted config dict, or path to a deepspeed config: https://www.deepspeed.ai/docs/config-json. All defaults will be ignored if a config is passed in. logging_level: Set logging level for deepspeed. loss_scale: Loss scaling value for FP16 training. 0.0 results in dynamic loss scaling, otherwise static. initial_scale_power: Power of the initial dynamic loss scale value. Loss scale is computed by ``2^initial_scale_power``. loss_scale_window: Window in which to raise/lower the dynamic FP16 loss scaling value. hysteresis: FP16 Delay shift in Dynamic Loss scaling. min_loss_scale: The minimum FP16 dynamic loss scaling value. partition_activations: Enables partition activation when used with ZeRO stage 3 and model parallelism. Still requires you to wrap your forward functions in deepspeed.checkpointing.checkpoint. See `deepspeed tutorial `_. cpu_checkpointing: Offloads partitioned activations to CPU if ``partition_activations`` is enabled. contiguous_memory_optimization: Copies partitioned activations so that they are contiguous in memory. Not supported by all models. synchronize_checkpoint_boundary: Insert :func:`torch.cuda.synchronize` at each checkpoint boundary. load_full_weights: True when loading a single checkpoint file containing the model state dict when using ZeRO Stage 3. This differs from the DeepSpeed checkpoint which contains shards per worker. ztTo use the `DeepSpeedStrategy`, you must have DeepSpeed installed. Install it by running `pip install -U deepspeed`.)r%rCrDrOrPN)r)r*r,r+r-r.r/r7r0r1r2r3r4r5r6rJrKrLrMr'r9r:r;r<r=r>r8rF)_DEEPSPEED_AVAILABLE ImportErrorsuper__init__Z_backward_sync_control _load_configrA_create_default_configr_config_initializedutilsloggingloggersetLevelr(rNrErFrGrHrI_deepspeed_engine).selfr%r&r'r(r)r*r+r,r-r.r/r0r1r2r3r4r5r6r7r8r9r:r;r<r=r>r?r@rArBrCrDrErFrGrHrIrJrKrLrMrNrOrPr __class__n/var/www/html/stable-diffusion-webui/venv/lib/python3.9/site-packages/lightning_fabric/strategies/deepspeed.pyrU/st0  !zDeepSpeedStrategy.__init__)rQcCs2t|jtsJ|jd}|duo0|ddkS)Nr&r') isinstancerAdictget)r^r&rararb zero_stage_3"s zDeepSpeedStrategy.zero_stage_3cCst|j|jdS)N)Z num_replicasrank)re world_size global_rankr^rararbdistributed_sampler_kwargs(sz,DeepSpeedStrategy.distributed_sampler_kwargszdeepspeed.DeepSpeedEnginecCs|jSN)r]rkrararbmodel,szDeepSpeedStrategy.model)module optimizersrQcCsJt|dkr tdt|d|||d\|_}||j|gfS)aSet up a model and multiple optimizers together. Currently, only a single optimizer is supported. Return: The model wrapped into a :class:`deepspeed.DeepSpeedEngine` and a list with a single deepspeed optimizer. r#z>Currently only one optimizer is supported with DeepSpeed. Got z optimizers instead.r)len ValueError_initialize_enginer]'_set_deepspeed_activation_checkpointing)r^rorp optimizerrararbsetup_module_and_optimizers0s z-DeepSpeedStrategy.setup_module_and_optimizers)rorQcCs||\|_}|jS)zwSet up a module for inference (no optimizers). For training, see :meth:`setup_module_and_optimizers`. )rsr])r^ro_rararb setup_moduleEszDeepSpeedStrategy.setup_module)rurQcCst|dS)zOptimizers can only be set up jointly with the model in this strategy. Please use :meth:`setup_module_and_optimizers` to set up both module and optimizer together. N)NotImplementedErrorZ_err_msg_joint_setup_required)r^rurararbsetup_optimizerMsz!DeepSpeedStrategy.setup_optimizer)NNNccsddl}|jr|jsJ|jjdkr,tj}n|jjdkr@tj}ntj}|jj |j d|j |ddVWdq1sz0YndVdS)Nr16bf16T)r(r7Zconfig_dict_or_pathdtype) rrgrXrOtorchfloat16Zbfloat16float32zeroZInitr(rA)r^rr}rararbmodule_sharded_contextTs    &z(DeepSpeedStrategy.module_sharded_context) checkpointfilepathstorage_optionsrQcCstdSrmry)r^rrrrararbsave_checkpointmsz!DeepSpeedStrategy.save_checkpoint)checkpoint_pathrQcCstdSrmr)r^rrararbload_checkpointpsz!DeepSpeedStrategy.load_checkpoint)rprrQcCsdSrmra)r^rprrararbload_optimizer_state_dictssz+DeepSpeedStrategy.load_optimizer_state_dict)rorrQcCs&|jr"|jr"|||||dSrm)rNrgZmodule_to_device_restore_zero_state)r^rorrararbload_module_state_dictys  z(DeepSpeedStrategy.load_module_state_dict)strategy_registryrQc Cs|jd|dd|jd|ddd|jd|d d d|jd |d d d d|jd|ddd|jd|ddd d d|jd|ddd d dddd dS)NrzDefault DeepSpeed Strategy) descriptionZdeepspeed_stage_1z#DeepSpeed with ZeRO Stage 1 enabledr#)rr'Zdeepspeed_stage_2z#DeepSpeed with ZeRO Stage 2 enabledrZdeepspeed_stage_2_offloadz&DeepSpeed ZeRO Stage 2 and CPU OffloadT)rr'r)Zdeepspeed_stage_3zDeepSpeed ZeRO Stage 3rcZdeepspeed_stage_3_offloadz&DeepSpeed ZeRO Stage 3 and CPU Offload)rr'r)r*Zdeepspeed_stage_3_offload_nvmez'DeepSpeed ZeRO Stage 3 and NVMe OffloadZnvme)rr'r)r*r(r+r0)register)clsrrararbregister_strategiess<z%DeepSpeedStrategy.register_strategies)rnrurQcCsNddl}tdd|}|jtj|jjd|j|||dd\}}}}||fS)zInitialize one model and one optimizer with an optional learning rate scheduler. This calls :func:`deepspeed.initialize` internally. rNcSs|jSrm)Z requires_grad)prararbz6DeepSpeedStrategy._initialize_engine..)Z device_rankF)argsrArnmodel_parametersruZdist_init_required) rfilter parametersZ initializeargparse NamespaceZ root_deviceindexrA)r^rnrurrZdeepspeed_engineZdeepspeed_optimizerrwrararbrss z$DeepSpeedStrategy._initialize_enginecCsXt|jts"td|jjjdt||jt _ | |j sT| d|_ dS)Nz;The DeepSpeed strategy is only supported on CUDA GPUs but `z ` is used.T)rdr%r RuntimeErrorr`__name__rZ_set_world_ranksrjrrh_init_deepspeed_distributedrX_format_configrkrararb_setup_distributeds z$DeepSpeedStrategy._setup_distributedcCsrddl}|jdusJtdkrP|td|jd|jdd|j| |_ |j |j |jj ddS)NrWindowsz1initializing deepspeed distributed: GLOBAL_RANK: z , MEMBER: r#/)Zdistributed_port) rrDplatformsystem_set_node_environment_variablesrinforjriZ_get_process_group_backendZ_process_group_backendZinit_distributed main_port)r^rrararbrs  z-DeepSpeedStrategy._init_deepspeed_distributedcCsb|jdusJ|jjtjd<t|jjtjd<t|jtjd<t|jtjd<t|jtjd<dS)NZ MASTER_ADDRZ MASTER_PORTZRANKZ WORLD_SIZEZ LOCAL_RANK) rDZ main_addressosenvironstrrrjriZ local_rankrkrararbrs z1DeepSpeedStrategy._set_node_environment_variablescCs`ddl}t|jtsJ|jdr\|jd}|jjd|d|d|d|dddS)Nractivation_checkpointingrJrLrKprofile)Zmpu_rJZcontiguous_checkpointingZcheckpoint_in_cpur)rrdrArerfZ checkpointing configure)r^rZcheckpoint_configrararbrts  z9DeepSpeedStrategy._set_deepspeed_activation_checkpointingcCs,|jdurtd|jdd|dS)NzTo use DeepSpeed you must pass in a DeepSpeed config dict, or a path to a JSON config. See: https://pytorch-lightning.readthedocs.io/en/stable/advanced/model_parallel.html#deepspeedtrain_micro_batch_size_per_gpur#)rArr setdefault_format_precision_configrkrararbrs  z DeepSpeedStrategy._format_configcCst|jtsJ|jjdkrRd|jvr~tdd|j|j|j|j|j d|jd<n,d|jvr~|jjdkr~tdddi|jd<dS) Nr{Zfp16zEnabling DeepSpeed FP16.T)enabledrErFrGrHrIr|zEnabling DeepSpeed BF16.r) rdrArerOrrErFrGrHrIrkrararbrs  z*DeepSpeedStrategy._format_precision_config)r&r?r@rJrKrLrMr)r*r,r+r-r.r/r0r1r7r2r3r4r5r6 zero_kwargsrQcKsx||||d|||||dd}|rh|}|r>|| ||d|d<| rX| | | | ||d|d<|||d|rt||d <|S) N)rJrKrLrM)r2r3r4r5r6)rZaio)devicer, buffer_countr7r))rr,r buffer_sizer/r7Z offload_param)r?r&r)update)r^r&r?r@rJrKrLrMr)r*r,r+r-r.r/r0r1r7r2r3r4r5r6rcfgZ zero_configrararbrWsH  z(DeepSpeedStrategy._create_default_config)rockptrQcs:ddldtjjtddfdd |dddS) a7Overrides the normal load_state_dict behaviour in PyTorch to ensure we gather parameters that may be sharded across processes before loading the state dictionary when using ZeRO stage 3. This is then automatically synced across processes. Args: ckpt: The ckpt file. rN)roprefixrQc sg}g}g}d}t|dd}|}|dur6||_|durBin||ddi}jjt|jdddd.jr|j |||d|||d Wdn1s0Y|j D]"\}} | dur| ||d qdS) N state_dict _metadataF)Zrecurser)Z modifier_rankT)rrlocal_metadatastrict missing_keysunexpected_keys error_msgs.) getattrcopyrrfrZGatheredParameterslistrZis_global_zeroZ_load_from_state_dictZ_modulesitems) rorrrrrmetadatarnamechildrrloadr^rarbrUs.  $ z3DeepSpeedStrategy._restore_zero_state..load)r)r)rr~nnrr)r^rorrarrbrKs" z%DeepSpeedStrategy._restore_zero_state)rArQcCs|dur2|jtjvr2td|jdtj|j}t|ttfrtj|sZt d|t |}t |}Wdn1s0Yt|t s|dusJ|S)Nz"Loading DeepSpeed config from set z environment variablezHYou passed in a path to a DeepSpeed config but the path does not exist: )DEEPSPEED_ENV_VARrrrrdrrpathisfileFileNotFoundErroropenjsonrre)r^rAfrararbrVws   (zDeepSpeedStrategy._load_config)N)N)6r __module__ __qualname__rrZWARNr rboolintrr rrrrr~rrfloatrrUpropertyrgrlrnrrr rvrxrzrrrrrrr rr classmethodrrsrrrrtrrrWrrV __classcell__rarar_rbr,s2 t  #   F,r)2rrrZrr contextlibrpathlibrtypingrrrrrr r r r r r~Z lightning_utilities.core.importsrZtorch.nnrZ torch.optimrZlightning_fabric.acceleratorsrrZ9lightning_fabric.plugins.environments.cluster_environmentrZ"lightning_fabric.plugins.precisionrZlightning_fabric.strategies.ddprZ$lightning_fabric.strategies.strategyrZ&lightning_fabric.utilities.distributedrZ$lightning_fabric.utilities.rank_zerorrZlightning_fabric.utilities.seedrZ lightning_fabric.utilities.typesrrRrrrarararbs0  0