hSrSSKJr SSKrSSKJr SSKrSSKrSSK J r J r J r SSK JrJr S S S jjrSS \ R"S4SS jjrg)zlFunctions for estimating the best YOLO batch size to use a fraction of the available CUDA memory in PyTorch.) annotationsN)deepcopy) DEFAULT_CFGLOGGERcolorstr)autocast profile_opsc[US9 [[U5R5USUs=:aS:aO OUOSUS9sSSS5 $!,(df  g=f)a Compute optimal YOLO training batch size using the autobatch() function. Args: model (torch.nn.Module): YOLO model to check batch size for. imgsz (int, optional): Image size used for training. amp (bool, optional): Use automatic mixed precision if True. batch (int | float, optional): Fraction of GPU memory to use. If -1, use default. max_num_obj (int, optional): The maximum number of objects from dataset. Returns: (int): Optimal batch size computed using the autobatch() function. Notes: If 0.0 < batch < 1.0, it's used as the fraction of GPU memory to use. Otherwise, a default fraction of 0.6 is used. )enabledgg?333333?)fraction max_num_objN)r autobatchrtrain)modelimgszampbatchrs U/home/james-whalen/.local/lib/python3.13/site-packages/ultralytics/utils/autobatch.pycheck_train_batch_sizersF0 #  UO ! ! #UcE>OC>OUUXfq    s 5A  Arc@[S5n[R"USUSUS-S35 [UR 55R nUR S;a[R"USU35 U$[RRR(a[R"USU35 U$S nS [R"S S 5R5S 3n[RR!U5n U R"U- n [RR%U5U- n [RR'U5U- n XU -- n [R"UUSU R(SU SSU SSU SSU SS3 5 U S:a/SQO/SQnUVs/sHn[R*"USX5PM nn[-UUSXdS9n[/[1UU55VVVs/sHvununnU(dM[3US[4[645(dM2S USs=:aU :dMDO MHUS :Xd!UUS- (aUSUUS- S:dMoUUS/PMx nnnnU(a[1U6O//4unn[8R:"UUSS9n[5[=X-5US- US - 5nSU;a+UR?S5nXU:aU[AUS- S 5nUS:dUS:a![R"USUS US!35 Un[8RB"UU5U -U -U - n[R"US"US#US$X-SS%U SS&US-S'S(3 5 U[RRE5 $s snfs snnnf![FaIn[R"US)US*US!35 UsSnA[RRE5 $SnAff=f![RRE5 f=f)+a Automatically estimate the best YOLO batch size to use a fraction of the available CUDA memory. Args: model (torch.nn.Module): YOLO model to compute batch size for. imgsz (int, optional): The image size used as input for the YOLO model. fraction (float, optional): The fraction of available CUDA memory to use. batch_size (int, optional): The default batch size to use if an error is detected. max_num_obj (int, optional): The maximum number of objects from dataset. Returns: (int): The optimal batch size. z AutoBatch: z'Computing optimal batch size for imgsz=z at dz% CUDA memory utilization.>cpumpsz4intended for CUDA devices, using default batch-size zHRequires torch.backends.cudnn.benchmark=False, using default batch-size i@zCUDA:CUDA_VISIBLE_DEVICES0rz (z) z.2fz G total, z G reserved, z G allocated, zG free)r r)r r r!r"r @r )ndevicerr )degNizbatch=z. outside safe range, using default batch-size .zUsing batch-size z for  zG/zG (z.0fu%) ✅zerror detected: z, using default batch-size )$rrinfonext parametersr'typewarningtorchbackendscudnn benchmarkosgetenvstripcudaget_device_properties total_memorymemory_reservedmemory_allocatednameemptyr enumeratezip isinstanceintfloatnppolyfitroundindexmaxpolyval empty_cache Exception)rrr batch_sizerprefixr'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 U A KK6(1#R 01S'1S'VWX[U\\ijkloippvwx'("f"2JK!8CD 1u{{1a. Dc5AfV 's;'@A A 6Aq 1Q4#u. AaD 1   awq1u~1Aq8I1I Q!IA  $&sBxB8 u JJue + q|$qt+qt3 4 7? d#AN"AE1 . q5AH NNfXVA3.\]g\hhij kAJJq!$q(1,1 vh/s%s!AL;MRPQRUwVYZbehZhilYmmstu   =E , &!1!4PQ[P\\]^_     sm1N'5"N'N'>N N 3N N  #N 0 N ;DN' N'' O:1 O5O:O=5O::O== P)r Tr ) rtorch.nn.ModulerrArboolrz int | floatrrAreturnrA) rrarrArrBrKrArrArcrA)__doc__ __future__rr4copyrnumpyrCr0ultralytics.utilsrrrultralytics.utils.torch_utilsrr rrrrrlsr"  ;;?              @!'' J! J! J!J! J!  J!  J!rk