o Ti_@sRddlZddlTGdddeZGdddeZGdd d eZGd d d ZdS) N)*c@eZdZdZdS)ElasticityErrorz: Base exception for all elasticity related errors N__name__ __module__ __qualname____doc__r r O/home/ubuntu/.local/lib/python3.10/site-packages/deepspeed/elasticity/config.pyr rc@r)ElasticityConfigErrorz( Elasticity configuration error Nrr r r r rr rc@r)ElasticityIncompatibleWorldSizezY Attempting to run a world size that is incompatible with a given elastic config Nrr r r r rr rc@s(eZdZdZddZddZddZdS) ElasticityConfiga Elastic config object, constructed from a param dictionary that only contains elastic config parameters, example below: If elasticity is enabled, user must specify (at least) max_train_batch_size and micro_batch_sizes. { "enabled": true, "max_train_batch_size": 2000, "micro_batch_sizes": [2,4,6], "min_gpus": 1, "max_gpus" : 10000 "min_time": 20 "ignore_non_elastic_batch_info": false "version": 0.1 } cCs|tt|_|jr,t|vr|t|_ntdtt|vr%|t|_ntdt|tt |_|tt |_t |jt sQtdtdt |jd|jttdd|jsftdtd|jttd d|js{tdtd |j|tt|_|tt|_|jd ks|jd krtd |jd |j|j|jkrtd|jd |j|tt|_|jd krtd|j|tt|_|jd krtd|j|tt|_|jdkrtd|j|tt |_!|t"t#|_$|t%t&|_'dS)NzElasticity config missing zElasticity expected value of z, to be a list of micro batches, instead is: z, containing: cSs t|tSN) isinstanceintmr r r Es z+ElasticityConfig.__init__..zElasticity expected z8 to only contain a list of integers, instead contains: fcSs|dkS)Nrr rr r r rIsz7 to only contain positive integers, instead contains: frz5Elasticity min/max gpus must be > 0, given min_gpus: z , max_gpus: zEElasticity min_gpus cannot be greater than max_gpus, given min_gpus: zFModel-Parallel size cannot be less than 1, given model-parallel size: zNNumber of GPUs per node cannot be less than 1, given number of GPUs per node: rz,Elasticity min time needs to be >= 0: given )(getENABLEDENABLED_DEFAULTenabledMAX_ACCEPTABLE_BATCH_SIZEmax_acceptable_batch_sizer MICRO_BATCHES micro_batches!MAX_ACCEPTABLE_BATCH_SIZE_DEFAULTMICRO_BATCHES_DEFAULTrlisttypeallmapMIN_GPUSMIN_GPUS_DEFAULTmin_gpusMAX_GPUSMAX_GPUS_DEFAULTmax_gpusMODEL_PARALLEL_SIZEMODEL_PARALLEL_SIZE_DEFAULTmodel_parallel_sizeNUM_GPUS_PER_NODENUM_GPUS_PER_NODE_DEFAULTnum_gpus_per_nodeMIN_TIMEMIN_TIME_DEFAULTmin_timeVERSIONVERSION_DEFAULTversionPREFER_LARGER_BATCHPREFER_LARGER_BATCH_DEFAULTprefer_larger_batch_sizeIGNORE_NON_ELASTIC_BATCH_INFO%IGNORE_NON_ELASTIC_BATCH_INFO_DEFAULTignore_non_elastic_batch_info)self param_dictr r r __init__0sx          zElasticityConfig.__init__cCs|jSr)__dict__r=r r r reprjszElasticityConfig.reprcCstj|jdddS)NT) sort_keysindent)jsondumpsr@rAr r r __repr__mszElasticityConfig.__repr__N)rrr r r?rBrHr r r r rs : r)rF constants Exceptionrrrrr r r r s