|h dZddlZddlmZddlmZddlZddlZddl m Z m Z m Z ddl mZmZ dd ej j"d ed ed eeefd edef dZdde j,dfd ej j"d ededed edef dZy)zlFunctions for estimating the best YOLO batch size to use a fraction of the available CUDA memory in PyTorch.N)deepcopy)Union) DEFAULT_CFGLOGGERcolorstr)autocast profile_opsmodelimgszampbatch max_num_objreturnct|5tt|j|d|cxkrdkrnn|nd|cdddS#1swYyxYw)a Compute optimal YOLO training batch size using the autobatch() function. Args: model (torch.nn.Module): YOLO model to check batch size for. imgsz (int, optional): Image size used for training. amp (bool, optional): Use automatic mixed precision if True. batch (int | float, optional): Fraction of GPU memory to use. If -1, use default. max_num_obj (int, optional): The maximum number of objects from dataset. Returns: (int): Optimal batch size computed using the autobatch() function. Notes: If 0.0 < batch < 1.0, it's used as the fraction of GPU memory to use. Otherwise, a default fraction of 0.6 is used. )enabledgg?333333?)fractionrN)r autobatchrtrain)r r rrrs Z/var/www/html/test/engine/venv/lib/python3.12/site-packages/ultralytics/utils/autobatch.pycheck_train_batch_sizersO0 #   UO ! ! #UcE>OC>OUUXfq    s 6A  Arr batch_sizectd}tj|d|d|dzdt|j j }|j dvrtj|d||Stjjjrtj|d||Sd }d tjd d jd }tjj!|} | j"|z } tjj%||z } tjj'||z } | | | zz } tj||d| j(d| dd| dd| dd| dd | dkrgdngd} |Dcgc]}tj*|d||}}t-||d||}t/t1||Dcgc]Y\}\}}|rOt3|dt4t6fr6d |dcxkr| kr(nn%|d k(s||dz r|d||dz dkDr||dg[}}}}|rt1|nggf\}}t9j:||d}t5t=| |z|dz |d z }d|vr+|j?d}|||k\r|tA|dz d }|dks|dkDr tj|d|d |d!|}t9jB||| z| z| z }tj|d"|d#|d$| |zdd%| dd&|dzd'd( |tjjEScc}wcc}}}w#tF$rH}tj|d)|d*|d!|cYd}~tjjESd}~wwxYw#tjjEwxYw)+a> Automatically estimate the best YOLO batch size to use a fraction of the available CUDA memory. Args: model (torch.nn.Module): YOLO model to compute batch size for. imgsz (int, optional): The image size used as input for the YOLO model. fraction (float, optional): The fraction of available CUDA memory to use. batch_size (int, optional): The default batch size to use if an error is detected. max_num_obj (int, optional): The maximum number of objects from dataset. Returns: (int): The optimal batch size. z AutoBatch: z'Computing optimal batch size for imgsz=z at dz% CUDA memory utilization.>cpumpsz4intended for CUDA devices, using default batch-size zHRequires torch.backends.cudnn.benchmark=False, using default batch-size i@zCUDA:CUDA_VISIBLE_DEVICES0rz (z) z.2fz G total, z G reserved, z G allocated, zG free)r r!)r r"r#r$r! @r )ndevicerr")degNizbatch=z. outside safe range, using default batch-size .zUsing batch-size z for  zG/zG (z.0fu%) ✅zerror detected: z, using default batch-size )$rrinfonext parametersr)typewarningtorchbackendscudnn benchmarkosgetenvstripcudaget_device_properties total_memorymemory_reservedmemory_allocatednameemptyr enumeratezip isinstanceintfloatnppolyfitroundindexmaxpolyval empty_cache Exception)r r rrrprefixr)gbd propertiestraf batch_sizesbimgresultsixyxyfit_xfit_ypes rrr-s*m $F KK6(A%XX[^L\\vwx %""$ % , ,F {{n$&!UV`Uabc ~~%%&!ijtiuvw B  0#6<<>qABCA11&9J"$A ""6*R/A ##F+b0A QU A KK6(1#R 01S'1S'VWX[U\\ijkloippvwx'("f"2JK!8CD1u{{1a.DDc5Af+V 's;'@A  6Aq1Q4#u.AaD 1 awq1u~1Aq8I1I !I  $&sBxB8 u JJue + q8|$qt+qt3 4 7? d#AKN"AE1 . q5AH NNfXVA3.\]g\hhij kAJJq!$q(1,1 vh/s%s!AL;MRPQRUwVYZbehZhilYmmstu   =E , &!1!4PQ[P\\]^_     sP%M5)M)+M53AM.C9M5) M55 O>OOO OO O))r Tr )__doc__r6copyrtypingrnumpyrEr2ultralytics.utilsrrrultralytics.utils.torch_utilsrr nnModulerCboolrDrrrrrmsr  ;;? !  88??     e       @!'' J! 88??J! J!J! J!  J!  J!rl