§
    l¬¸j, ã                  ó*  — d Z ddlmZ ddlZddlZddlZddlZddlZddlZddl	m	Z	m
Z
 ddlmZmZ ddlmZ ddlmZ ddlZddlZddlmZ dd	lmZmZ dd
lmZ ddlmZmZmZ ddlmZm Z m!Z! ddl"m#Z# ddl$m%Z% ddl&m'Z' ddl(m)Z)m*Z*m+Z+m,Z,m-Z-m.Z.m/Z/m0Z0m1Z1m2Z2m3Z3m4Z4 ddl5m6Z6 ddl7m8Z8m9Z9m:Z:m;Z;m<Z< ddl=m>Z>m?Z? ddl@mAZA ddlBmCZC ddlDmEZE ddlFmGZGmHZHmIZImJZJmKZKmLZLmMZMmNZNmOZOmPZPmQZQmRZRmSZSmTZTmUZUmVZVmWZWmXZXmYZYmZZZm[Z[m\Z\  G d„ d¦  «        Z] G d„ d¦  «        Z^dS )zu
Train a model on a dataset.

Usage:
    $ yolo train model=yolo26n.pt data=coco8.yaml imgsz=640 epochs=100 batch=16
é    )ÚannotationsN)ÚcopyÚdeepcopy)ÚdatetimeÚ	timedelta)Úpartial)ÚPath)Údistributed)ÚnnÚoptim)Ú__version__)Ú_YOLO_CLI_COMMANDÚget_cfgÚget_save_dir)Úcheck_cls_datasetÚcheck_det_datasetÚ convert_ndjson_to_yolo_if_needed)ÚDistillationModel)Úload_checkpoint)ÚMuSGD)ÚDEFAULT_CFGÚGITÚ
LOCAL_RANKÚLOGGERÚRANKÚTQDMÚWINDOWSÚYAMLÚ	callbacksÚ	clean_urlÚcolorstrÚemojis)Úcheck_train_batch_size)Ú	check_ampÚ
check_fileÚcheck_imgszÚcheck_model_file_from_stemÚ
print_args)Úddp_cleanupÚgenerate_ddp_command)Úget_latest_run)Úoverride_configs)Úplot_results)Ú
TORCH_1_11Ú	TORCH_2_0Ú	TORCH_2_4Ú
TORCH_2_13ÚEarlyStoppingÚModelEMAÚattempt_compileÚautocastÚ$convert_optimizer_state_dict_to_fp16Úget_torch_device_backendÚ
init_seedsÚ	one_cycleÚparse_deviceÚprepare_qatÚ	qat_stateÚrestore_qatÚselect_deviceÚstrip_optimizerÚ	strip_qatÚtorch_distributed_zero_firstÚunset_deterministicÚunwrap_modelc                  ó8  — e Zd ZdZeddfd>d„Zd?d„Zd?d	„Zd?d
„Zd„ Z	d„ Z
d„ Zd„ Zd„ Zd„ Zd„ Zd@d„ZdAd„ZdBdCd„Zd„ Zd„ Zd„ Zd„ Zd„ Zd„ Zd„ Zd „ ZdDd"„Zd#„ ZdEd&„ZdFd'„ZdGd(„Zd)„ Z d*„ Z!d+„ Z"d,„ Z#d-„ Z$d.„ Z%d/„ Z&d0„ Z'dBd1„Z(d2„ Z)d3„ Z*d4„ Z+d5„ Z,d6„ Z-d7„ Z.dHd=„Z/dS )IÚBaseTraineraw	  A base class for creating trainers.

    This class provides the foundation for training YOLO models, handling the training loop, validation, checkpointing,
    and various training utilities. It supports both single-GPU and multi-GPU distributed training.

    Attributes:
        args (SimpleNamespace): Configuration for the trainer.
        validator (BaseValidator): Validator instance.
        model (nn.Module): Model instance.
        callbacks (defaultdict): Dictionary of callbacks.
        save_dir (Path): Directory to save results.
        wdir (Path): Directory to save weights.
        last (Path): Path to the last checkpoint.
        best (Path): Path to the best checkpoint.
        save_period (int): Save checkpoint every x epochs (disabled if < 1).
        batch_size (int): Batch size for training.
        epochs (int): Number of epochs to train for.
        start_epoch (int): Starting epoch for training.
        device (torch.device): Device to use for training.
        amp (bool): Whether Automatic Mixed Precision is enabled.
        scaler (torch.amp.GradScaler): Gradient scaler for AMP.
        data (dict): Dataset dictionary containing paths and metadata.
        ema (ModelEMA): EMA (Exponential Moving Average) of the model.
        resume (bool): Resume training from a checkpoint.
        lf (callable): Learning rate scheduling function.
        scheduler (torch.optim.lr_scheduler._LRScheduler): Learning rate scheduler.
        best_fitness (float): The best fitness value achieved.
        fitness (float): Current fitness value.
        loss (torch.Tensor): Current loss value.
        tloss (dict): Running mean of loss items.
        loss_names (tuple): Names of loss items, derived from the loss dict returned by the criterion on the first
            batch.
        csv (Path): Path to results CSV file.
        metrics (dict): Dictionary of metrics.
        plots (dict): Dictionary of plots.

    Methods:
        train: Execute the training process.
        validate: Run validation on the val set.
        save_model: Save model training checkpoints.
        get_dataset: Get train and validation datasets.
        setup_model: Load, create, or download model.
        build_optimizer: Construct an optimizer for the model.

    Examples:
        Initialize a trainer and start training
        >>> trainer = BaseTrainer(cfg="config.yaml")
        >>> trainer.train()
    NÚ
_callbacksúdict | Nonec                ó&	  ‡— t          ||¦  «        | _        |                      |pi ¦  «         t          | j        dd¦  «        rKt	          | j        j        d         t          ¦  «        s&ddlŠˆfd„| j        j        D ¦   «         | j        _        t          | j        j	        ¦  «        | j        _	        t          | j        j	        ¦  «        | _	        | j	        j        dvrt          | j	        ¦  «        nd| _        d| _        d| _        i | _        t#          | j        j        dz   t&          z   | j        j        ¬¦  «         t+          | j        ¦  «        | _        | j        j        | j        _        | j        dz  | _        t&          d	v ri| j                             d
d
¬¦  «         t5          | j        ¦  «        | j        _        t7          j        | j        dz  t;          | j        ¦  «        ¦  «         | j        dz  | j        dz  c| _        | _        | j        j         | _         | j        j!        | _"        | j        j#        | _#        d| _$        t&          dk    r!tK          t;          | j        ¦  «        ¦  «         | j	        j        dv rd| j        _&        |rtO          |¦  «        ntQ          j)        ¦   «         | _(        | j(         *                    ¦   «         D ]!\  }}| '                    ¦   «         | j(        |<   Œ"| j	        j        dv rd}n:| j        j	        r,tW          | j        j	         ,                    d¦  «        ¦  «        nd}|dk    o
tZ          dk    | _.        || _/        t&          d	v r0| j.        s)tQ          j0        | ¦  «         |  1                    d¦  «         te          | j        j3        ¦  «        | _3        ti          tZ          ¦  «        5  |  5                    ¦   «         | _6        ddd¦  «         n# 1 swxY w Y   d| _7        d| _8        d| _9        d| _:        d| _;        d| _<        d| _=        d| _>        d| _?        | j        dz  | _@        | j@         A                    ¦   «         r%| j        jB        s| j@         C                    ¦   «          g d¢| _D        d| _E        dS )a5  Initialize the BaseTrainer class.

        Args:
            cfg (str | dict | SimpleNamespace, optional): Path to a configuration file or configuration object.
            overrides (dict, optional): Configuration overrides.
            _callbacks (dict, optional): Dictionary of callback functions.
        ÚaugmentationsNr   c                ó:   •— g | ]}‰                      |¦  «        ‘ŒS © )Úto_dict)Ú.0ÚtÚAs     €új/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/ultralytics/engine/trainer.pyú
<listcomp>z(BaseTrainer.__init__.<locals>.<listcomp>�   s#   ø€ Ð&UÐ&UÐ&U¸ q§y¢y°¡|¤|Ð&UÐ&UÐ&Uó    >   ÚcpuÚmpsé   )ÚdeterministicÚweights¾   r   éÿÿÿÿT©ÚparentsÚexist_okz	args.yamlúlast.ptúbest.ptrY   ú,Úon_pretrain_routine_startrK   zresults.csv)r   rU   é   )Fr   ÚargsÚcheck_resumeÚgetattrÚ
isinstancerI   ÚdictÚalbumentationsr:   Údevicer>   Útyper7   ÚacceleratorÚ	validatorÚmetricsÚplotsr8   Úseedr   rV   r   Úsave_dirÚnameÚwdirÚmkdirÚstrr   ÚsaveÚvarsÚlastÚbestÚsave_periodÚbatchÚ
batch_sizeÚepochsÚstart_epochr(   Úworkersr   r   Úget_default_callbacksÚitemsÚlenÚsplitr   ÚddpÚ
world_sizeÚadd_integration_callbacksÚrun_callbacksr'   ÚmodelrA   Úget_datasetÚdataÚemaÚlfÚ	optimizerÚ	schedulerÚbest_fitnessÚfitnessÚlossÚtlossÚ
loss_namesÚcsvÚexistsÚresumeÚunlinkÚplot_idxÚnan_recovery_attempts)ÚselfÚcfgÚ	overridesrF   ÚkÚvrƒ   rO   s          @rP   Ú__init__zBaseTrainer.__init__‚   sK  ø€ õ ˜C Ñ+Ô+ˆŒ	Ø×Ò˜)˜/ rÑ*Ô*Ð*Ý�4”9˜o¨tÑ4Ô4ð 	V½ZÈÌ	ÔH_Ð`aÔHbÕdhÑ=iÔ=ið 	VØ&Ð&Ð&Ð&à&UÐ&UÐ&UÐ&U¸T¼YÔ=TÐ&UÑ&UÔ&UˆDŒIÔ#Ý'¨¬	Ô(8Ñ9Ô9ˆŒ	ÔÝ# D¤IÔ$4Ñ5Ô5ˆŒØDHÄKÔDTÐ\jÐDjÐDjÕ3°D´KÑ@Ô@Ð@ÐptˆÔØˆŒØˆŒØˆŒ
Ý�4”9”> AÑ%­Ñ,¸D¼IÔ<SÐTÑTÔTÐTõ % T¤YÑ/Ô/ˆŒØœÔ+ˆŒ	ŒØ”M IÑ-ˆŒ	Ý�7ˆ?ˆ?ØŒI�OŠO D°4ˆOÑ8Ô8Ð8Ý!$ T¤]Ñ!3Ô!3ˆDŒIÔÝŒI�d”m kÑ1µ4¸¼	±?´?ÑCÔCÐCØ#œy¨9Ñ4°d´iÀ)Ñ6KÐˆŒ	�4”9Øœ9Ô0ˆÔàœ)œ/ˆŒØ”iÔ&ˆŒØˆÔÝ�2Š:ˆ:Ý•t˜DœI‘”Ñ'Ô'Ð'ð Œ;Ô˜~Ð-Ð-Ø !ˆDŒIÔð .8Ð^�˜jÑ)Ô)Ð)½YÔ=\Ñ=^Ô=^ˆŒØ”N×(Ò(Ñ*Ô*ð 	)ð 	)‰DˆAˆqØ !§¢¡¤ˆDŒN˜1ÑÐð Œ;Ô˜~Ð-Ð-ØˆJˆJà=A¼YÔ=MÐT�˜TœYÔ-×3Ò3°CÑ8Ô8Ñ9Ô9Ð9ÐSTˆJà ’>Ð6¥j°BÒ&6ˆŒØ$ˆŒå�7ˆ?ˆ? 4¤8ˆ?ÝÔ/°Ñ5Ô5Ð5Ø×ÒÐ:Ñ;Ô;Ð;õ 0°´	´Ñ@Ô@ˆŒ
Ý)­*Ñ5Ô5ð 	+ð 	+Ø×(Ò(Ñ*Ô*ˆDŒIð	+ð 	+ð 	+ñ 	+ô 	+ð 	+ð 	+ð 	+ð 	+ð 	+ð 	+øøøð 	+ð 	+ð 	+ð 	+ð ˆŒð ˆŒØˆŒØˆŒð !ˆÔØˆŒØˆŒ	ØˆŒ
ØˆŒØ”= =Ñ0ˆŒØŒ8�?Š?ÑÔð 	 T¤YÔ%5ð 	ØŒH�OŠOÑÔÐØ!˜	˜	ˆŒØ%&ˆÔ"Ð"Ð"s   ÏO*Ï*O.Ï1O.Úeventrs   c                óF   — | j         |                              |¦  «         dS )z7Append the given callback to the event's callback list.N)r   Úappend©r˜   rž   Úcallbacks      rP   Úadd_callbackzBaseTrainer.add_callback×   s#   € àŒ�uÔ×$Ò$ XÑ.Ô.Ð.Ð.Ð.rR   c                ó   — |g| j         |<   dS )zPOverride the existing callbacks with the given callback for the specified event.N)r   r¡   s      rP   Úset_callbackzBaseTrainer.set_callbackÛ   s   € à!) 
ˆŒ�uÑÐÐrR   c                óX   — | j                              |g ¦  «        D ]} || ¦  «         ŒdS )z>Run all existing callbacks associated with a particular event.N)r   Úgetr¡   s      rP   r…   zBaseTrainer.run_callbacksß   s<   € àœ×*Ò*¨5°"Ñ5Ô5ð 	ð 	ˆHØˆH�T‰NŒNˆNˆNð	ð 	rR   c                óÖ  — 	 | j         �r| j        j        r t          j        d¦  «         d| j        _        | j        j        dk     r#t          d| j        › d| j        dz  › d�¦  «        ‚d\  }}	 t          | ¦  «        \  }}t          j	        t          d	¦  «        › d
d                     |¦  «        › �¦  «         t          j        |d¬¦  «         |�t          | t          |¦  «        ¦  «         n9# |�t          | t          |¦  «        ¦  «         w w xY w|                      ¦   «          t#          ¦   «          n# t#          ¦   «          w xY w| j         s|                      d¦  «         dS dS )zcExecute the training process, using DDP subprocess for multi-GPU or direct training for single-GPU.zI'rect=True' is incompatible with Multi-GPU training, setting 'rect=False'Fç      ð?zuAutoBatch with batch<1 not supported for Multi-GPU training, please specify a valid batch size multiple of GPU count z, i.e. batch=é   ú.©NNzDDP:z debug command ú T©ÚcheckNÚteardown)r‚   rb   Úrectr   Úwarningry   Ú
ValueErrorrƒ   r*   Úinfor!   ÚjoinÚ
subprocessÚrunr)   rs   Ú	_do_trainrB   r…   )r˜   ÚcmdÚfiles      rP   ÚtrainzBaseTrainer.trainä   s©  € ð	"ØŒxñ !à”9”>ð +Ý”NÐ#nÑoÔoÐoØ%*�D”I”NØ”9”? SÒ(Ð(Ý$ðHØSWÔSbðHð HØquô  rAð  DEñ  rEðHð Hð Hñô ð ð '‘	��Tð5Ý 4°TÑ :Ô :‘I�C˜Ý”K¥8¨FÑ#3Ô#3Ð SÐ SÀCÇHÂHÈSÁMÄMÐ SÐ SÑTÔTÐTÝ”N 3¨dÐ3Ñ3Ô3Ð3àÐ'Ý# D­#¨d©)¬)Ñ4Ô4Ð4øøð Ð'Ý# D­#¨d©)¬)Ñ4Ô4Ð4Ð4ð (øøøð —’Ñ Ô Ð åÑ!Ô!Ð!Ð!øÕÑ!Ô!Ð!Ð!øøøØŒxð 	+Ø×Ò˜zÑ*Ô*Ð*Ð*Ð*ð	+ð 	+s%   ‚A,D8 Á/A!C0 Ã D8 Ã0"DÄD8 Ä8Ec                óà   ‡ — ‰ j         j        r&t          d‰ j         j        ‰ j        ¦  «        ‰ _        n
ˆ fd„‰ _        t          j                             ‰ j	        ‰ j        ¬¦  «        ‰ _
        dS )z,Initialize training learning rate scheduler.rU   c                ót   •— t          d| ‰j        z  z
  d¦  «        d‰j        j        z
  z  ‰j        j        z   S )NrU   r   r©   )Úmaxr{   rb   Úlrf)Úxr˜   s    €rP   ú<lambda>z.BaseTrainer._setup_scheduler.<locals>.<lambda>	  s6   ø€ ¥ A¨¨D¬K©Ñ$7¸Ñ ;Ô ;¸sÀTÄYÄ]Ñ?RÑ SÐVZÔV_ÔVcÑ c€ rR   )Ú	lr_lambdaN)rb   Úcos_lrr9   r¿   r{   rŠ   r   Úlr_schedulerÚLambdaLRr‹   rŒ   ©r˜   s   `rP   Ú_setup_schedulerzBaseTrainer._setup_scheduler  sa   ø€ àŒ9Ôð 	dÝ  4¤9¤=°$´+Ñ>Ô>ˆDŒGˆGàcÐcÐcÐcˆDŒGÝÔ+×4Ò4°T´^ÈtÌwÐ4ÑWÔWˆŒˆˆrR   c                ó–   — t          | j        j        t          | j        dz
  d¦  «        ¦  «        }|dk    rt          ||z  ¦  «        ndS )zPReturn warmup iterations, leaving at least the final epoch for regular training.rU   r   )Úminrb   Úwarmup_epochsr¾   r{   Úround)r˜   Únum_batchesrÊ   s      rP   Ú_get_warmup_iterationsz"BaseTrainer._get_warmup_iterations  sH   € å˜DœIÔ3µS¸¼Àq¹È!Ñ5LÔ5LÑMÔMˆØ5BÀQÒ5FÐ5F�u�] [Ñ0Ñ1Ô1Ð1ÈAÐMrR   c                ó   — | j         j                             dd¦  «        d         }|dv r|nd}| j         j                             dd¦  «        d                              d¦  «        }t          |t                   ¦  «        }t          j        ||¦  «        | _        t          | j        ¦  «        | _        | j                             |¦  «         |dk    rdt          j
        d	<   n=|d
k    r7t          t          d¦  «        rt          j        ¦   «         st          d¦  «        ‚t          j        dddœ                     |t          j        ¦   «         rdnd¦  «        t%          d¬¦  «        t&          | j        ¬¦  «         dS )zGInitialize and set the DistributedDataParallel parameters for training.ú:rU   r   >   ÚnpuÚxpuÚcudarY   r_   Ú1ÚTORCH_NCCL_BLOCKING_WAITrÑ   Úis_xccl_availablezOMulti-XPU training requires XCCL, which is not available in this PyTorch build.ÚhcclÚxccl)rÐ   rÑ   ÚncclÚglooi0*  )Úseconds)ÚbackendÚtimeoutÚrankrƒ   N)rb   rh   r�   Úintr   Útorchr7   rj   Ú
set_deviceÚosÚenvironÚhasattrÚdistrÕ   ÚRuntimeErrorÚinit_process_groupr§   Úis_nccl_availabler   r   rƒ   )r˜   Údevice_typeÚdevicesÚindexs       rP   Ú
_setup_ddpzBaseTrainer._setup_ddp  sk  € à”iÔ&×,Ò,¨S°!Ñ4Ô4°QÔ7ˆØ%0°NÐ%BÐ%B�k�kÈˆØ”)Ô"×(Ò(¨¨aÑ0Ô0°Ô4×:Ò:¸3Ñ?Ô?ˆÝ�G�JÔ'Ñ(Ô(ˆÝ”l ;°Ñ6Ô6ˆŒÝ3°D´KÑ@Ô@ˆÔØÔ×#Ò# EÑ*Ô*Ð*Ø˜&Ò Ð Ø58�BŒJÐ1Ñ2Ð2Ø˜EÒ!Ð!­7µ4Ð9LÑ+MÔ+MÐ!ÕRVÔRhÑRjÔRjÐ!ÝÐpÑqÔqÐqÝÔØ"¨6Ð2Ð2×6Ò6°{ÍdÔNdÑNfÔNfÐDrÀFÀFÐlrÑsÔsÝ eÐ,Ñ,Ô,ÝØ”ð		
ñ 	
ô 	
ð 	
ð 	
ð 	
rR   c                ó®  — | j         t          | j        d¦  «        z  }|                      | j        d         |t
          d¬¦  «        | _        t          | j        j        ¦  «        | j        j         z  p| j        j         }| j	        j
        d| j        z  k     r9| j        j        s-|dk    r't          d| j	        j
        › dd| j        z  › �¦  «        ‚|                      | j        | j	        j                 | j	        j        dv r|n|dz  t
          d¬¦  «        | _        t          t#          | j	        j        | j         z  ¦  «        d¦  «        | _        | j	        j        | j         z  | j        z  | j	        j        z  }| j        €¬t-          j        t          | j        j        ¦  «        t          | j         | j	        j        ¦  «        z  ¦  «        | j        z  }|                      | j        | j	        j        | j	        j        | j	        j        ||¬
¦  «        | _        |                      ¦   «          d	S | j        j        D ]}|                      d¦  «        dv r||d<   Œd	S )zKBuild dataloaders and update optimizer settings for the current batch size.rU   r»   ©rz   rÝ   Úmodera   z final batch=1 training at imgsz=zJ gives BatchNorm a single value per channel; change batch or use imgsz >= >   ÚobbÚdepthÚsemanticÚvalN)r†   rp   ÚlrÚmomentumÚdecayÚ
iterationsÚparam_group>   ÚmuonÚweightÚweight_decay)!rz   r¾   rƒ   Úget_dataloaderrˆ   r   Útrain_loaderr€   Úsamplerrb   ÚimgszÚstrideÚ	drop_lastr³   r�   ÚtaskÚtest_loaderrË   ÚnbsÚ
accumulaterú   r‹   ÚmathÚceilÚdatasetr{   Úbuild_optimizerr†   Úlr0rô   rÇ   Úparam_groupsr§   )r˜   rz   Úfinal_batch_sizerú   rö   Úgroups         rP   Ú_build_train_pipelinez!BaseTrainer._build_train_pipeline%  sX  € à”_­¨D¬O¸QÑ(?Ô(?Ñ?ˆ
Ø ×/Ò/ØŒI�gÔ¨:½JÈWð 0ñ 
ô 
ˆÔõ ˜tÔ0Ô8Ñ9Ô9¸DÔ<MÔ<XÑXÐxÐ\`Ô\mÔ\xÐØŒ9Œ?˜Q ¤™_Ò,Ð,°TÔ5FÔ5PÐ,ÐUeÐijÒUjÐUjÝðB°4´9´?ð Bð BØ01°D´K±ðBð Bñô ð ð
  ×.Ò.ØŒI�d”i”oÔ&Ø%)¤Y¤^Ð7SÐ%SÐ%S�z�zÐYcÐfgÑYgÝØð	 /ñ 
ô 
ˆÔõ �e D¤I¤M°D´OÑ$CÑDÔDÀaÑHÔHˆŒØ”yÔ-°´Ñ?À$Ä/ÑQÐTXÔT]ÔTaÑaˆØŒ>Ð!Ýœ¥3 tÔ'8Ô'@Ñ#AÔ#AÅCÈÌÐY]ÔYbÔYfÑDgÔDgÑ#gÑhÔhÐkoÔkvÑvˆJØ!×1Ò1Ø”jØ”YÔ(Ø”9”=ØœÔ+Ø"Ø%ð 2ñ ô ˆDŒNð ×!Ò!Ñ#Ô#Ð#Ð#Ð#àœÔ4ð 9ð 9�Ø—9’9˜]Ñ+Ô+Ð/AÐAÐAØ,8�E˜.Ñ)øð9ð 9rR   c           	     óÒ  ‡— |                       ¦   «         }| j                             | j        ¦  «        | _        | j        j        du p| j        j        du ot          ot           }|r;| j        j        dk    r+| j                             t          j        ¬¦  «        | _        n.| j        j        r"t          j        d| j        j        › d�¦  «         |                      ¦   «          t          t          t          | j        d¦  «        r| j        j                             ¦   «         nd¦  «        d¦  «        }t#          | j        j        ||d	¬
¦  «        | j        _        || _        | j        j        dk    �r| j        r1|                     d¦  «        rt-          | j        |d         ¦  «         nØ| j        d	k    r+t          | j        t          | j        d	¦  «        z  d	¦  «        nd}t3          t4          ¦  «        5  t7          | j        ddi¦  «        5  |                      | j        d         |dd¬¦  «        }ddd¦  «         n# 1 swxY w Y   ddd¦  «         n# 1 swxY w Y   t=          | j        || j        ¦  «        | _        ~| j        j         �,| j        j!        r t          j        d¦  «         d| j        _!        tE          | j        | j        | j        j!        ¬¦  «        | _        tG          | j        j$        tJ          ¦  «        r| j        j$        n9tG          | j        j$        t          ¦  «        rtM          | j        j$        ¦  «        ng }dg}d„ |D ¦   «         |z   }tG          tO          | j        ¦  «        tP          ¦  «        r| )                    d¦  «         || _*        | j         +                    ¦   «         D ]r\  Š}	tY          ˆfd„|D ¦   «         ¦  «        r t          j-        d‰› d�¦  «         d|	_.        Œ@|	j.        s+|	j/        j0        rt          j        d‰› d�¦  «         d|	_.        ŒstY          d„ | j         1                    ¦   «         D ¦   «         ¦  «        ste          d| j        j$        › d�¦  «        ‚| j        j3        d v| _3        t          j4        | j3        ¦  «                             | j        ¦  «        | _3        | j3        ru| j        j3        d!k    retj          d"v r\tl          j7         8                    ¦   «         }
t          j4        ts          | j        ¦  «        | j        ¬#¦  «        | _3        |
tl          _7        tj          dk    rD| j        d	k    r9| j3                             ¦   «         | _3        tu          j;        | j3        d$¬%¦  «         ty          | j3        ¦  «        | _3        | j        j        d&k    r?d$dl=}|j>        j3         ?                    | j3        o| j        j3        d!k    ¬'¦  «        | _@        n˜t‚          rRt          j3         ?                    | j        j        d(k    r| j        j        nd| j3        o| j        j3        d!k    ¬'¦  «        n9t          jB        j3         ?                    | j3        o| j        j3        d!k    ¬'¦  «        | _@        | j        j         �LtG          tO          | j        ¦  «        tP          ¦  «        s%tQ          | j        | j        j         ¬)¦  «        | _        | j        d	k    ryt          rd*ty          | j        j!        ¦  «        ini }d|t†          rd+nd,<   t‰          jE        jF        | j        f| j        jG        gty          | j        j!        ¦  «         d-œ|¤Ž| _        | j        d	k     r0tj          dk    r%|  H                    ¦   «         x| j        _I        | _        |  J                    ¦   «          |  K                    ¦   «         | _L        |  M                    ¦   «          t�          | j        ¦  «        | _O        tj          d"v r{| jL        jP        jQ        |  R                    d.¬/¦  «        z   }t§          t©          |d$gt«          |¦  «        z  ¦  «        ¦  «        | _P        | j        jV        r|  W                    ¦   «          t±          | j        jY        ¬0¦  «        dc| _Z        | _[        |  \                    |¦  «         | j]        d	z
  | j^        __        |  `                    d1¦  «         dS )2zYConfigure model, optimizer, dataloaders, and training utilities before the training loop.TNrÒ   ©Úmemory_formatz='channels_last=True' is only supported on CUDA, ignoring on 'z'.rÿ   é    rU   )rÿ   ÚfloorÚmax_dimrª   Úmodelopté   ÚcacheFr»   rY   rí   zL'compile' is not supported with knowledge distillation and will be disabled.)rh   rî   z.dflc                ó   — g | ]}d |› d�‘Œ	S )zmodel.r«   rK   ©rM   rÀ   s     rP   rQ   z,BaseTrainer._setup_train.<locals>.<listcomp>{  s    € ÐAÐAÐA°˜m q˜m˜m˜mÐAÐAÐArR   zteacher_model.c              3  ó    •K  — | ]}|‰v V — Œ	d S ©NrK   )rM   rÀ   r›   s     €rP   ú	<genexpr>z+BaseTrainer._setup_train.<locals>.<genexpr>�  s'   øè è € Ð6Ð6˜a�1˜�6Ð6Ð6Ð6Ð6Ð6Ð6rR   zFreezing layer 'ú'z/setting 'requires_grad=True' for frozen layer 'zE'. See ultralytics.engine.trainer for customization of frozen layers.c              3  ó$   K  — | ]}|j         V — Œd S r  )Úrequires_grad©rM   rœ   s     rP   r  z+BaseTrainer._setup_train.<locals>.<genexpr>Š  s$   è è € ÐDÐD q�1”?ÐDÐDÐDÐDÐDÐDrR   z'freeze=zu' froze the entire model with no trainable parameters left. Reduce 'freeze' or pass a list of specific layer indices.>   FÚfp32Úbf16rX   ©rh   r   ©ÚsrcrÐ   )ÚenabledrÑ   ©Ústudent_modelÚteacher_modelÚstatic_graphÚforward_sync_buffersÚbroadcast_buffers)Ú
device_idsÚfind_unused_parametersrò   ©Úprefix)ÚpatienceÚon_pretrain_routine_end)aÚsetup_modelr†   Útorh   rb   Úchannels_lastr.   r   ri   rß   r   r²   Úset_model_attributesr¾   rÞ   rã   rÿ   r&   rþ   Úquantizer”   r§   r=   rz   rƒ   rA   r   r,   rû   rˆ   r;   Úpreprocess_batchÚdistill_modelÚcompiler4   re   ÚfreezeÚlistÚrangerC   r   r    Úfreeze_layer_namesÚnamed_parametersÚanyr´   r  ÚdtypeÚis_floating_pointÚ
parametersrå   ÚampÚtensorr   r   Údefault_callbacksr   r$   rä   Ú	broadcastÚboolÚ	torch_npurÐ   Ú
GradScalerÚscalerr0   rÒ   r1   r   ÚparallelÚDistributedDataParallelrê   Ú
auto_batchry   r  Úget_validatorrk   Úset_class_weightsr3   r‰   rl   ÚkeysÚlabel_loss_itemsrf   Úzipr€   rm   Úplot_training_labelsr2   r0  ÚstopperÚstopÚresume_trainingr|   rŒ   Ú
last_epochr…   )r˜   Úckptr4  Úgsry   Úcalibration_loaderÚfreeze_listÚalways_freeze_namesr=  rœ   Úcallbacks_backuprH  Ú
ddp_kwargsÚmetric_keysr›   s                 @rP   Ú_setup_trainzBaseTrainer._setup_trainJ  sÒ  ø€ à×ÒÑ!Ô!ˆØ”Z—]’] 4¤;Ñ/Ô/ˆŒ
ð œ	Ô/°4Ð7ð 
ØŒIÔ# tÐ+ÐJµ
ÐJÅ7¸{ð 	ð ð 	q˜Tœ[Ô-°Ò7Ð7ØœŸšµUÔ5H˜ÑIÔIˆDŒJˆJØŒYÔ$ð 	qÝŒNÐoÐ[_Ô[fÔ[kÐoÐoÐoÑpÔpÐpØ×!Ò!Ñ#Ô#Ð#õ •µ¸¼
ÀHÑ0MÔ0MÐU�T”ZÔ&×*Ò*Ñ,Ô,Ð,ÐSUÑVÔVÐXZÑ[Ô[ˆÝ% d¤i¤o¸bÈÐTUÐVÑVÔVˆŒ	ŒØˆŒð Œ9Ô Ò"Ñ"ØŒ{ð 	'˜tŸxšx¨
Ñ3Ô3ð 	'Ý˜DœJ¨¨ZÔ(8Ñ9Ô9Ð9Ð9àNRÌoÐabÒNbÐNb�˜DœO­s°4´?ÀAÑ/FÔ/FÑFÈÑJÔJÐJÐhj�Ý1µ*Ñ=Ô=ð ð Õ?OÐPTÔPYÐ\cÐejÐ[kÑ?lÔ?lð ð Ø)-×)<Ò)<Øœ	 'Ô*°uÀ2ÈGð *=ñ *ô *Ð&ðð ð ñ ô ð ð ð ð ð ð øøøð ð ð ð ð ð ð ñ ô ð ð ð ð ð ð øøøð ð ð ð õ )¨¬Ð5GÈÔI^Ñ_Ô_�”
Ø&ð Œ9Ô"Ð.°4´9Ô3DÐ.ÝŒNÐiÑjÔjÐjØ %ˆDŒIÔÝ$ T¤Z¸¼È$Ì)ÔJ[Ð\Ñ\Ô\ˆŒ
õ
 ˜$œ)Ô*­DÑ1Ô1ðˆDŒIÔÐõ ˜$œ)Ô*­CÑ0Ô0ð•�t”yÔ'Ñ(Ô(Ð(àð 	ð  &˜hÐØAÐA°[ÐAÑAÔAÐDWÑWÐÝ•l 4¤:Ñ.Ô.Õ0AÑBÔBð 	8Ø×%Ò%Ð&6Ñ7Ô7Ð7Ø"4ˆÔØ”J×/Ò/Ñ1Ô1ð 
	'ð 
	'‰DˆAˆqåÐ6Ð6Ð6Ð6Ð#5Ð6Ñ6Ô6Ñ6Ô6ð 'Ý”Ð3¨qÐ3Ð3Ð3Ñ4Ô4Ð4Ø"'�”�Ø”_ð '¨¬Ô)Bð 'Ý”ðYÀað Yð Yð Yñô ð ð #'�”øÝÐDÐD¨D¬J×,AÒ,AÑ,CÔ,CÐDÑDÔDÑDÔDð 	ÝðM˜4œ9Ô+ð Mð Mð Mñô ð ð ”9”=¨Ð7ˆŒÝ”< ¤Ñ)Ô)×,Ò,¨T¬[Ñ9Ô9ˆŒØŒ8ð 	;˜œ	œ¨Ò/Ð/µD¸G°O°OÝ(Ô:×?Ò?ÑAÔAÐÝ”|¥I¨d¬jÑ$9Ô$9À$Ä+ÐNÑNÔNˆDŒHØ*:�IÔ'Ý�"Š9ˆ9˜œ¨1Ò,Ð,Ø”x—|’|‘~”~ˆDŒHÝŒN˜4œ8¨Ð+Ñ+Ô+Ð+Ý˜œ‘>”>ˆŒØŒ;Ô˜uÒ$Ð$ØÐÐÐà#œ-Ô+×6Ò6¸t¼xÐ?cÈDÌIÌMÐ]cÒLcÐ6ÑdÔdˆDŒKˆKõ ð	]•”	×$Ò$Ø(,¬Ô(8¸EÒ(AÐ(A�D”KÔ$Ð$ÀvØ œHÐ@¨¬¬¸&Ò)@ð %ñ ô ð õ
 ”Z”^×.Ò.°t´xÐ7[ÀDÄIÄMÐU[ÒD[Ð.Ñ\Ô\ð ŒKð Œ9Ô"Ð.µzÅ,ÈtÌzÑBZÔBZÕ\mÑ7nÔ7nÐ.Ý*¸¼ÐSWÔS\ÔSjÐkÑkÔkˆDŒJØŒ?˜QÒÐõ GQÐX˜.­$¨t¬yÔ/@Ñ*AÔ*AÐBÐBÐVXˆJØX]ˆJµÐTÐ-Ð-ÐATÑUÝœÔ<Ø”
ðà œKÔ-Ð.Ý+/°´	Ô0AÑ+BÔ+BÐ'Bðð ð ð	ð ˆDŒJð Œ?˜QÒÐ¥4¨2¢: :Ø04·²Ñ0AÔ0AÐAˆDŒIŒO˜dœoØ×"Ò"Ñ$Ô$Ð$Ø×+Ò+Ñ-Ô-ˆŒØ×ÒÑ Ô Ð Ý˜DœJÑ'Ô'ˆŒÝ�7ˆ?ˆ?Øœ.Ô0Ô5¸×8MÒ8MÐUZÐ8MÑ8[Ô8[Ñ[ˆKÝ¥ K°!°µs¸;Ñ7GÔ7GÑ1GÑ HÔ HÑIÔIˆDŒLØŒyŒð ,Ø×)Ò)Ñ+Ô+Ð+å"/¸¼Ô9KÐ"LÑ"LÔ"LÈeÐˆŒ�d”iØ×Ò˜TÑ"Ô"Ð"Ø$(Ô$4°qÑ$8ˆŒÔ!Ø×ÒÐ4Ñ5Ô5Ð5Ð5Ð5s6   Ç7IÈ%I È4IÉ I	ÉIÉI	ÉIÉIÉIc                ó   ‡ ‡‡— ‰ j         dk    r‰                      ¦   «          ‰                      ¦   «          t          ‰ j        ¦  «        }‰                      |¦  «        }d}d‰ _        t          j        ¦   «         ‰ _        t          j        ¦   «         ‰ _	        ‰  
                    d¦  «         t          j        dt          ‰ j        j        ¦  «        › dt          ‰ j        j        ¦  «        › d‰ j        j        › d‰ j        j        › d	‰ j        j        ‰ j         pdz  › d
t'          d‰ j        ¦  «        › d�‰ j        j        r‰ j        j        › d�n	‰ j        › d�z   ¦  «         ‰ j        j        r:‰ j        ‰ j        j        z
  |z  }‰ j                             ||dz   |dz   g¦  «         ‰ j        }‰ j        j         }‰ j                             ¦   «          d‰ _        	 |‰ _        ‰  
                    d¦  «         t=          j        ¦   «         5  t=          j         d¦  «         ‰ j!         "                    ¦   «          ddd¦  «         n# 1 swxY w Y   ‰  #                    ¦   «          tH          dk    r‰ j        j%         &                    |¦  «         tO          ‰ j        ¦  «        }|sG|‰ j        ‰ j        j        z
  k    r/d}‰  (                    ¦   «          ‰ j         )                    ¦   «          tH          dv rP‰ j*        r&t          j        ‰  +                    ¦   «         ¦  «         tY          tO          ‰ j        ¦  «        |¬¦  «        }d‰ _-        |D �]\  Š}‰  
                    d¦  «         ‰||z  z   }	|	|k     �rd|g}
t]          dt_          ta          j1        |	|
d‰ j        j2        ‰ j3        z  g¦  «         4                    ¦   «         ¦  «        ¦  «        ‰ _5        ‰ j        j6        D ]«}to          ta          j1        |	|
| 8                    d¦  «        dk    r‰ j        j9        nd|d         ‰  :                    |¦  «        z  g¦  «        ¦  «        |d<   d|v r<to          ta          j1        |	|
‰ j        j;        ‰ j        j<        g¦  «        ¦  «        |d<   Œ¬	 t{          ‰ j        j>        dk    rt~          j@        n‰ j>        ‰ jA        jB        ¬¦  «        5  ‰  C                    |¦  «        }‰ j        jD        rL‰  E                    |d         ¦  «        }t�          ‰ jE        ¦  «         G                    ||¦  «        \  }‰ _H        n‰  E                    |¦  «        \  }‰ _H        | I                    ¦   «         ‰ _G        tH          dk    r‰ xjG        ‰ j         z  c_G        ‰ j*        s�t•          ‰ jH        ¦  «        ‰ _*        tH          dv rmt          j        ‰  +                    ¦   «         ¦  «         ‰ jK         L                    tš           N                    ‰  O                    d ¬!¦  «        d¦  «        ¦  «         ‰ j-        €‰ jH        n%ˆˆ fd"„‰ jH         P                    ¦   «         D ¦   «         ‰ _-        ddd¦  «         n# 1 swxY w Y   ‰ jQ         R                    ‰ jG        ¦  «         S                    ¦   «          �nó# t¨          $ �råŠd#t«          ‰¦  «         V                    ¦   «         v }|st¯          ˆfd$„d%D ¦   «         ¦  «        s‚ |‰ j        k    s‰ j        d&k    stH          dk    r‚ ‰ xj        dz  c_        ‰ j3        }t]          ‰ j3        dz  d¦  «        x‰ j        _X        ‰ _3        |r!‰ jA        jB         Y                    ¦   «         › d'�nd(}t          jZ        |› d)|› d*‰ j3        › d+‰ j        › d,�¦  «         dx}x}}dx‰ _G        x‰ _H        ‰ _-        t·          ‰ j        d-¦  «        r‰ j         \                    ¦   «          ‰  ]                    ¦   «          ‰  ^                    ¦   «          ‰ j        j         }‰ j        ‰ j_        _`        ‰ j        dz
  ‰ j!        _a        t          ‰ j        ¦  «        }‰                      |¦  «        }d}‰ j                             ¦   «          Y dŠ‹ �nÓdŠ‹ww xY w|	|z
  ‰ j5        k    rŸ‰  b                    ¦   «          |	}‰ j        j        r}t          j        ¦   «         ‰ j	        z
  ‰ j        j        d.z  k    ‰ _c        tH          dk    r7tH          dk    r‰ jc        ndg}tÉ          je        |d¦  «         |d         ‰ _c        ‰ jc        r �ntH          dv rît          ‰ j-        ¦  «        }| f                    d/d0d|z   z  z   |dz   › d1‰ j        › �‰  g                    ¦   «         d2›d3�g‰ j-         h                    ¦   «         ¢| 8                    d4|d         ¦  «        ji        d         ‘|d         ji        d         ‘R z  ¦  «         ‰  
                    d5¦  «         ‰ j        jj        r|	‰ j        v r‰  k                    ||	¦  «         ‰  
                    d6¦  «         ‰ jc        r n	�Œd‰ _        ‰ j        r	‰ jc        s�Œœt·          t�          ‰ jE        ¦  «        jl        d7¦  «        r+t�          ‰ jE        ¦  «        jl         L                    ¦   «          d8„ tO          ‰ j        j6        ¦  «        D ¦   «         ‰ _m        ‰  
                    d9¦  «         tH          dv r#‰ jn         o                    ‰ jE        g d:¢¬;¦  «         |dz   ‰ j        k    }‰ j        jp        s|s‰ jq        jr        s‰ jc        rH‰  ]                    ‰ jA        jB        d<k    rdnd=¦  «         ‰  s                    ¦   «         \  ‰ _K        ‰ _t        ‰  u                    |¦  «        r�	Œåd‰ _v        tH          dv ré‰  w                    i ‰  O                    ‰ j-        ¦  «        ¥‰ jK        ¥‰ jm        ¥¬>¦  «         ‰ xjc        ‰  q                    |dz   ‰ jt        ¦  «        p|z  c_c        ‰ j        j        r:‰ xjc        t          j        ¦   «         ‰ j	        z
  ‰ j        j        d.z  k    z  c_c        ‰ j        jx        s|r)‰  y                    ¦   «         r‰  
                    d?¦  «         t          j        ¦   «         }|‰ j        z
  ‰ _        |‰ _        ‰ j        j        r |‰ j	        z
  |‰ j        z
  dz   z  }tõ          j{        ‰ j        j        d.z  |z  ¦  «        x‰ _        ‰ j        _        ‰                      |¦  «        }‰  |                    ¦   «          ‰ j        ‰ j!        _a        ‰ xjc        |‰ j        k    z  c_c        ‰  
                    d@¦  «         ‰  ]                    ‰ jA        jB        d<k    rdnd=¦  «         tH          dk    r7tH          dk    r‰ jc        ndg}tÉ          je        |d¦  «         |d         ‰ _c        ‰ jc        rn|dz  }�Œ?t          j        ¦   «         ‰ j	        z
  }t          j        dA|‰ j        z
  dz   › dB|d.z  dC›dD�¦  «         ‰  }                    ¦   «          tH          dv r5‰ j        jj        r‰  ~                    ¦   «          ‰  
                    dE¦  «         ‰  ]                    ¦   «          ‰ j        ‰ j        fD ]&}t·          |d-¦  «        r| \                    ¦   «          Œ'dS )FzbPerform the full training loop including setup, epoch iteration, validation, and final evaluation.rU   rY   NÚon_train_startzUsing z train, z val images for fraction=z
 at imgsz=z
Using z' dataloader workers
Logging results to Úboldz
Starting training for z	 hours...z
 epochs...ra   r   TÚon_train_epoch_startÚignorerX   )ÚtotalÚon_train_batch_startr÷   Úbiasç        Ú
initial_lrró   rô   r!  r"  Úimgrò   r.  c                óJ   •— i | ]\  }}|‰j         |         ‰z  |z   ‰d z   z  “Œ S )rU   )r�   )rM   r›   rœ   Úir˜   s      €€rP   ú
<dictcomp>z)BaseTrainer._do_train.<locals>.<dictcomp>%  s8   ø€ Ð!kÐ!kÐ!kÉ4È1Èa ! d¤j°¤m°aÑ&7¸!Ñ&;ÀÀAÁÑ%FÐ!kÐ!kÐ!krR   zout of memoryc              3  ó:   •K  — | ]}|t          ‰¦  «        v V — Œd S r  )rs   )rM   ÚsÚes     €rP   r  z(BaseTrainer._do_train.<locals>.<genexpr>,  sA   øè è € ð .ð .àð �S ™VœV˜ð.ð .ð .ð .ð .ð .rR   )ÚCUBLAS_STATUS_ALLOC_FAILEDÚCUDNN_STATUS_INTERNAL_ERRORzunable to find an engineé   z out of memoryzCUDA backend memory errorz with batch=z. Reducing to batch=z and retrying (z/3).Úclosei  z%11s%11sz%11.4gú/z.3gÚGÚclsÚon_batch_endÚon_train_batch_endÚupdatec                ó,   — i | ]\  }}d |› �|d         “ŒS )zlr/pgró   rK   )rM   ÚirrÀ   s      rP   rn  z)BaseTrainer._do_train.<locals>.<dictcomp>y  s(   € ÐaÐaÐa±°°Q�|˜r�|�| Q t¤WÐaÐaÐarR   Úon_train_epoch_end)ÚyamlÚncrb   Únamesrÿ   Úclass_weights)ÚincluderT   g      à?)rl   Úon_model_saveÚon_fit_epoch_endú
z epochs completed in z.3fz hours.Úon_train_end)rƒ   rë   r`  r€   rü   rÍ   Ú
epoch_timeÚtimeÚepoch_time_startÚtrain_time_startr…   r   r´   r  r  rb   Úfractionrþ   Únum_workersr!   ro   r{   Úclose_mosaicr–   Úextendr|   r‹   Ú	zero_gradÚ_oom_retriesÚepochÚwarningsÚcatch_warningsÚsimplefilterrŒ   ÚstepÚ_model_trainr   rý   Ú	set_epochÚ	enumerateÚ_close_dataloader_mosaicÚresetr‘   Úprogress_stringr   r�   r¾   rÞ   ÚnpÚinterpr  rz   rË   r  r
  Úfloatr§   Úwarmup_bias_lrrŠ   Úwarmup_momentumrô   r5   rC  rß   Úbfloat16rh   ri   r7  r9  r†   rC   r�   Ú
loss_itemsÚsumÚtuplerl   r{  rf   ÚfromkeysrQ  r   rJ  ÚscaleÚbackwardrå   rs   Úlowerr?  ry   Úupperr²   rã   ru  Ú_clear_memoryr  rk   Ú
dataloaderrW  Úoptimizer_steprU  rä   Úbroadcast_object_listÚset_descriptionÚ_get_memoryÚvaluesÚshaperm   Úplot_training_samplesÚ	criterionró   r‰   Úupdate_attrrò   rT  Úpossible_stopÚvalidaterŽ   Ú_handle_nan_recoveryr—   Úsave_metricsrt   Ú
save_modelr  r  rÇ   Ú
final_evalÚplot_metrics)r˜   ÚnbÚnwÚlast_opt_stepÚbase_idxr’  Úmosaic_closedÚpbarry   ÚniÚxirÀ   Úpredsr�   Úis_oomÚ	old_batchÚerrorÚbroadcast_listÚloss_lengthÚfinal_epochrN   Úmean_epoch_timerÚ   Úloaderrq  rm  s   `                       @@rP   r¸   zBaseTrainer._do_trainÉ  s¦  øøø€ àŒ?˜QÒÐØ�OŠOÑÔÐØ×ÒÑÔÐå�Ô"Ñ#Ô#ˆØ×(Ò(¨Ñ,Ô,ˆØˆØˆŒÝ $¤	¡¤ˆÔÝ $¤	¡¤ˆÔØ×ÒÐ+Ñ,Ô,Ð,ÝŒð&•S˜Ô*Ô2Ñ3Ô3ð &ð &½SÀÔAQÔAYÑ=ZÔ=Zð &ð &Øœ	Ô*ð&ð &Ø6:´i´oð&ð &àÔ&Ô2°d´oÐ6JÈÑKð&ð &õ #+¨6°4´=Ñ"AÔ"Að&ð &ð &ð JNÌÌÐ)w¨D¬I¬NÐ)EÐ)EÐ)EÐ)EÐ`dÔ`kÐ]wÐ]wÐ]wñ	yñ	
ô 	
ð 	
ð Œ9Ô!ð 	IØœ d¤iÔ&<Ñ<ÀÑBˆHØŒM× Ò  (¨H°q©L¸(ÀQ¹,Ð!GÑHÔHÐHØÔ ˆØ œIÔ2Ð2ˆØŒ× Ò Ñ"Ô"Ð"ØˆÔðF	ØˆDŒJØ×ÒÐ5Ñ6Ô6Ð6ÝÔ(Ñ*Ô*ð &ð &ÝÔ% hÑ/Ô/Ð/Ø”×#Ò#Ñ%Ô%Ð%ð&ð &ð &ñ &ô &ð &ð &ð &ð &ð &ð &øøøð &ð &ð &ð &ð ×ÒÑÔÐÝ�rŠzˆzØÔ!Ô)×3Ò3°EÑ:Ô:Ð:Ý˜TÔ.Ñ/Ô/ˆDà ð * U¨d¬k¸D¼IÔ<RÑ.RÒ%RÐ%RØ $�Ø×-Ò-Ñ/Ô/Ð/ØÔ!×'Ò'Ñ)Ô)Ð)å�wˆˆØ”?ð 8Ý”K × 4Ò 4Ñ 6Ô 6Ñ7Ô7Ð7Ý�I dÔ&7Ñ8Ô8ÀÐCÑCÔC�ØˆDŒJØ ð w&ñ w&‘��5Ø×"Ò"Ð#9Ñ:Ô:Ð:à˜˜e™‘^�Ø˜’7‘7Ø˜R˜�BÝ&)¨!­Sµ´¸2¸rÀAÀtÄyÄ}ÐW[ÔWfÑGfÐCgÑ1hÔ1h×1nÒ1nÑ1pÔ1pÑ-qÔ-qÑ&rÔ&r�D”OØ!œ^Ô8ð vð v˜å"'ÝœIØ "Ø "à@AÇÂÀmÑ@TÔ@TÐX^Ò@^Ð@^ D¤IÔ$<Ð$<ÐdgØ$% l¤O°d·g²g¸e±n´nÑ$Dð!"ñô ñ	#ô 	#˜˜$™ð &¨˜?˜?Ý,1µ"´)¸BÀÀTÄYÔE^Ð`dÔ`iÔ`rÐDsÑ2tÔ2tÑ,uÔ,u˜A˜j™Møð;Ý!°D´I´MÀVÒ4KÐ4K¥%¤. .ÐQUÔQYÐbfÔbmÔbrÐsÑsÔsð ð Ø $× 5Ò 5°eÑ <Ô <˜Øœ9Ô,ð Fà$(§J¢J¨u°U¬|Ñ$<Ô$<˜EÝ4@ÀÄÑ4LÔ4L×4QÒ4QÐRWÐY^Ñ4_Ô4_Ñ1˜D $¤/ /à48·J²J¸uÑ4EÔ4EÑ1˜D $¤/Ø$(§H¢H¡J¤J˜œ	Ý 2š:˜:Ø ˜IœI¨¬Ñ8˜IœIØ#œð mÝ.3°D´OÑ.DÔ.D˜DœOÝ# w˜˜Ý &¤¨D×,@Ò,@Ñ,BÔ,BÑ CÔ CÐ CØ $¤× 3Ò 3µD·M²MÀ$×BWÒBWÐ_dÐBWÑBeÔBeÐgjÑ4kÔ4kÑ lÔ lÐ lð  $œzÐ1ð !œO˜Oà!kÐ!kÐ!kÐ!kÐ!kÐSWÔSb×ShÒShÑSjÔSjÐ!kÑ!kÔ!kð œ
ð!ð ð ñ ô ð ð ð ð ð ð øøøð ð ð ð ð. ”K×%Ò% d¤iÑ0Ô0×9Ò9Ñ;Ô;Ð;Ñ;øÝ#ð "ñ "ð "Ø,µ°A±´·²±´Ð>�FØ!ð ­#ð .ð .ð .ð .ð"ð.ñ .ô .ñ +ô +ð ð Ø˜tÔ/Ò/Ð/°4Ô3DÈÒ3IÐ3IÍTÐUWÊZÈZØØÐ%Ô%¨Ñ*Ð%Ô%Ø $¤�IÝ8;¸D¼OÈqÑ<PÐRSÑ8TÔ8TÐT�D”I”O d¤oØKQÐr˜tœ{Ô/×5Ò5Ñ7Ô7ÐGÐGÐGÐGÐWr�EÝ”NØ ð eð e¨ið eð eØ-1¬_ðeð eØMQÔM^ðeð eð eñô ð ð ,0Ð/�EÐ/˜D 5Ø?CÐC�D”IÐC ¤°$´*Ý˜tÔ0°'Ñ:Ô:ð 2ØÔ)×/Ò/Ñ1Ô1Ð1Ø×&Ò&Ñ(Ô(Ð(Ø×.Ò.Ñ0Ô0Ð0Ø(,¬	Ô(>Ð$>�MØ04Ô0@�D”NÔ-Ø04Ô0@À1Ñ0D�D”NÔ-Ý˜TÔ.Ñ/Ô/�BØ×4Ò4°RÑ8Ô8�BØ$&�MØ”N×,Ò,Ñ.Ô.Ð.Ø�E�E�E�E‘EøøøøðE"øøøðF ˜Ñ%¨¬Ò8Ð8Ø×'Ò'Ñ)Ô)Ð)Ø$&�Mð ”y”~ð "Ý%)¤Y¡[¤[°4Ô3HÑ%HÈTÌYÌ^Ð^bÑMbÒ$c˜œ	Ý 2š:˜:Ý;?À1º9¸9¨d¬i¨iÈ$Ð-O˜NÝ Ô6°~ÀqÑIÔIÐIØ(6°qÔ(9˜DœIØœ9ð "Ø!™Eõ ˜7�?�?Ý"% d¤j¡/¤/�KØ×(Ò(Ø# h°!°k±/Ñ&BÑBà$ q™yÐ8Ð8¨4¬;Ð8Ð8Ø#×/Ò/Ñ1Ô1Ð8Ð8Ð8Ð8ðð "œZ×.Ò.Ñ0Ô0ðð "ŸIšI e¨U°5¬\Ñ:Ô:Ô@ÀÔCð	ð
 " %œLÔ.¨rÔ2ðð ññ	ô 	ð 	ð ×&Ò& ~Ñ6Ô6Ð6Ø”y”ð >¨2°´Ð+>Ð+>Ø×2Ò2°5¸"Ñ=Ô=Ð=à×"Ò"Ð#7Ñ8Ô8Ð8Ø”9ð Ø�Eñð %&�Ô!àÔ ð ¨¬ð Ùå•| D¤JÑ/Ô/Ô9¸8ÑDÔDð <Ý˜TœZÑ(Ô(Ô2×9Ò9Ñ;Ô;Ð;àaÐa½)ÀDÄNÔD_Ñ:`Ô:`ÐaÑaÔaˆDŒGà×ÒÐ3Ñ4Ô4Ð4Ý�wˆˆØ”×$Ò$ T¤ZÐ9sÐ9sÐ9sÐ$ÑtÔtÐtð   !™) t¤{Ò2ˆKØŒyŒ}ð = ð =¨t¬|Ô/Ið =ÈTÌYð =Ø×"Ò"¨4¬;Ô+;¸uÒ+DÐ+D 4 4È#ÑNÔNÐNØ-1¯]ª]©_¬_Ñ*�”˜dœlð ×(Ò(¨Ñ/Ô/ð Ùà)*ˆDÔ&Ý�wˆˆØ×!Ò!Ð*j¨T×-BÒ-BÀ4Ä:Ñ-NÔ-NÐ*jÐRVÔR^Ð*jÐbfÔbiÐ*jÐ!ÑkÔkÐkØ�	”	˜TŸ\š\¨%°!©)°T´\ÑBÔBÐQÀkÑQ�	”	Ø”9”>ð aØ�I”I¥$¤)¡+¤+°Ô0EÑ"EÈ$Ì)Ì.Ð[_ÑJ_Ò!`Ñ`�I”Ið ”I”Nð 8 kð 8°t·²Ñ7HÔ7Hð 8Ø×&Ò& Ñ7Ô7Ð7õ ”	‘”ˆAØ $Ô"7Ñ7ˆDŒOØ$%ˆDÔ!ØŒyŒ~ð 2Ø#$ tÔ'<Ñ#<ÀÈÔIYÑAYÐ\]ÑA]Ñ"^�Ý15´¸4¼9¼>ÈDÑ;PÐSbÑ;bÑ1cÔ1cÐc�”˜dœiÔ.Ø×0Ò0°Ñ4Ô4�Ø×%Ò%Ñ'Ô'Ð'Ø,0¬J�”Ô)Ø�	”	˜U d¤kÒ1Ñ1�	”	Ø×ÒÐ1Ñ2Ô2Ð2à×Ò t¤{Ô'7¸5Ò'@Ð'@˜t˜tÀcÑJÔJÐJõ �rŠzˆzÝ/3°qªy¨y $¤) )¸dÐ!C�ÝÔ*¨>¸1Ñ=Ô=Ð=Ø*¨1Ô-�”	ØŒyð ØØ�Q‰JˆEñMF	õP ”)‘+”+ Ô 5Ñ5ˆÝŒÐg˜ Ô!1Ñ1°AÑ5ÐgÐgÈGÐVZÉNÐgÐgÐgÐgÑhÔhÐhà�ŠÑÔÐÝ�7ˆ?ˆ?ØŒyŒð $Ø×!Ò!Ñ#Ô#Ð#Ø×Ò˜~Ñ.Ô.Ð.Ø×ÒÑÔÐØÔ(¨$Ô*:Ð;ð 	ð 	ˆFÝ�v˜wÑ'Ô'ð Ø—’‘”�øð	ð 	sP   È.IÉIÉIÒ<Z,ÓFY.Ù"Z,Ù.Y2	Ù2Z,Ù5Y2	Ù64Z,Ú,bÚ7Gbâbr   c                óþ   — t          j        | j        j        d| j        j        z   z  | j        z  ¦  «        | j        z  }t          | j        || j        j        dk    rt          j
        n| j        | j        ||¬¦  «        S )zJCalculate optimal batch size based on model and device memory constraints.rU   r!  )r†   rþ   rC  ry   Úmax_num_objÚdataset_size)r  r  rb   rþ   Úmulti_scalerÿ   r#   r†   rC  rß   r¢  rz   )r˜   rÏ  rÐ  Ú	max_imgszs       rP   rM  zBaseTrainer.auto_batch¹  sy   € õ ”I˜dœiœo°°T´YÔ5JÑ1JÑKÈdÌkÑYÑZÔZÐ]aÔ]hÑhˆ	Ý%Ø”*ØØ"&¤)¤-°6Ò"9Ð"9•”�¸t¼xØ”/Ø#Ø%ð
ñ 
ô 
ð 	
rR   Fc                ó†  — d\  }}| j         j        dk    rJt          j                             ¦   «         }|r)t          d¦  «                             ¦   «         j        dz  S nO| j         j        dk    r?| j         	                    ¦   «         }|r$| j         
                    | j         ¦  «        j        }|r|dk    r||z  ndn|dz  S )zJGet accelerator memory utilization in GB or as a fraction of total memory.©r   r   rT   Úpsutiléd   rS   r   i   @)rh   ri   rß   rT   Údriver_allocated_memoryÚ
__import__Úvirtual_memoryÚpercentrj   Úmemory_reservedÚget_device_propertiesÚtotal_memory)r˜   rŒ  Úmemoryrf  s       rP   r°  zBaseTrainer._get_memoryÆ  sË   € à‰ˆ�ØŒ;Ô˜uÒ$Ð$Ý”Y×6Ò6Ñ8Ô8ˆFØð KÝ! (Ñ+Ô+×:Ò:Ñ<Ô<ÔDÀsÑJÐJðKàŒ[Ô Ò&Ð&ØÔ%×5Ò5Ñ7Ô7ˆFØð YØÔ(×>Ò>¸t¼{ÑKÔKÔX�Ø9AÐW E¨A¢I I�˜%‘�°1°1ÈÐQVÉÐWrR   Ú	thresholdúfloat | Nonec                óP  — |r5d|cxk    rdk    sn J d¦   «         ‚|                       d¬¦  «        |k    rdS t          j        ¦   «          | j        j        dk    r t
          j                             ¦   «          dS | j        j        dk    rdS | j                             ¦   «          dS )	zIClear accelerator memory by calling garbage collector and emptying cache.r   rU   z"Threshold must be between 0 and 1.T)rŒ  NrT   rS   )	r°  ÚgcÚcollectrh   ri   rß   rT   Úempty_cacherj   )r˜   rß  s     rP   r«  zBaseTrainer._clear_memoryÓ  s¾   € àð 	Ø˜	Ð&Ð&Ò&Ð& QÒ&Ð&Ð&Ð&Ð&Ð(LÑ&Ô&Ð&Ø×Ò¨ÐÑ.Ô.°)Ò;Ð;Ø�Ý
Œ
‰ŒˆØŒ;Ô˜uÒ$Ð$ÝŒI×!Ò!Ñ#Ô#Ð#Ð#Ð#ØŒ[Ô Ò&Ð&ØˆFàÔ×(Ò(Ñ*Ô*Ð*Ð*Ð*rR   c                ó´   — ddl }	 |                     | j                             ¦   «         d¬¦  «                             d¬¦  «        S # t
          $ r i cY S w xY w)z0Read results.csv into a dictionary using polars.r   N)Úinfer_schema_lengthF)Ú	as_series)ÚpolarsÚread_csvr’   Ú
read_bytesrL   Ú	Exception)r˜   Úpls     rP   Úread_results_csvzBaseTrainer.read_results_csvá  sm   € àÐÐÐð	Ø—;’;˜tœx×2Ò2Ñ4Ô4È$�;ÑOÔO×WÒWÐbgÐWÑhÔhÐhøÝð 	ð 	ð 	ØˆIˆIˆIð	øøøs   †AA ÁAÁAc                ó  ‡— | j                              ¦   «          | j                              ¦   «         D ]X\  Š}t          t	          ˆfd„| j        ¦  «        ¦  «        r.t          |t          j        ¦  «        r| 	                    ¦   «          ŒYdS )zSet model in training mode.c                ó   •— | ‰v S r  rK   )ÚfÚns    €rP   rÁ   z*BaseTrainer._model_train.<locals>.<lambda>ï  s   ø€  A¨ F€ rR   N)
r†   r»   Únamed_modulesr?  Úfilterr=  re   r   ÚBatchNorm2dÚeval)r˜   Úmrñ  s     @rP   r—  zBaseTrainer._model_trainê  s�   ø€ àŒ
×ÒÑÔÐà”J×,Ò,Ñ.Ô.ð 	ð 	‰DˆAˆqÝ•6Ð*Ð*Ð*Ð*¨DÔ,CÑDÔDÑEÔEð Í*ÐUVÕXZÔXfÑJgÔJgð Ø—’‘”�øð	ð 	rR   c           
     óP  — ddl }| j        j        }t          d„ |                     ¦   «                              ¦   «         D ¦   «         ¦  «        sÙt          | j        ¦  «                             ¦   «         }|                     ¦   «                              ¦   «         D ]Œ\  }}t          |t          j
        ¦  «        rmt          j        |¦  «                             ¦   «         sGt          j        ||         ¦  «                             ¦   «         r|                     ||         ¦  «         Œ�t          |¦  «        }t          |¦  «                             ¦   «                              t          j        ¬¦  «        }t%          |d¦  «        rd|_        |                     ¦   «                              ¦   «         D ]D}t          |t          j
        ¦  «        r(|                     ¦   «         rt          j        |¦  «         ŒEt-          |¦  «         |                     ¦   «         }t          j        i d| j        “d| j        “dd“d	|“d
| j        j        “d|“dt9          t          | j                             ¦   «         ¦  «        ¦  «        “d| j                             ¦   «         “dt?          | j         ¦  «        “di | j!        ¥d| j"        i¥“d|  #                    ¦   «         “dtI          j%        ¦   «          &                    ¦   «          '                    ¦   «         “dtP          “dtS          tT          j+        ¦  «        tT          j,        tT          j-        tT          j.        tT          j/        dœ“dd“dd“|¦  «         | 0                    ¦   «         }| j1         2                    dd¬¦  «         | j3         4                    |¦  «         | j        | j"        k    r| j5         4                    |¦  «         | j6        dk    r9| j        | j6        z  dk    r&| j1        d| j        › d�z   4                    |¦  «         dS )z9Save model training checkpoints with additional metadata.r   Nc              3  ó–   K  — | ]D}t          |t          j        ¦  «        ¯t          j        |¦  «                             ¦   «         V — ŒEd S r  ©re   rß   ÚTensorÚisfiniteÚallr  s     rP   r  z)BaseTrainer.save_model.<locals>.<genexpr>û  sM   è è € ÐmÐm¨qÕQ[Ð\]Õ_dÔ_kÑQlÔQlÐm•5”> !Ñ$Ô$×(Ò(Ñ*Ô*ÐmÐmÐmÐmÐmÐmrR   r  r´  r’  r�   r†   r‰   Úupdatesr  r‹   rJ  Ú
train_argsÚtrain_metricsrŽ   Útrain_resultsÚdateÚversionÚgit)ÚrootÚbranchÚcommitÚmessageÚoriginÚlicensez*AGPL-3.0 (https://ultralytics.com/license)Údocszhttps://docs.ultralytics.comTrZ   ú.pt)7Úior‰   rü  Ú
state_dictr±  rC   r†   r   re   rß   rú  rû  Úcopy_r<   r   Úhalfr3  Úcontiguous_formatrã   r´  rA  Únan_to_num_r@   ÚBytesIOrt   r’  r�   rý  r6   r‹   rJ  ru   rb   rl   rŽ   rí  r   ÚnowÚ
astimezoneÚ	isoformatr   rs   r   r  r  r  r  r  Úgetvaluerq   rr   rv   Úwrite_bytesrw   rx   )	r˜   r  r‰   Úmodel_sdr›   rœ   r  ÚbufferÚserialized_ckpts	            rP   rº  zBaseTrainer.save_modelò  sý  € àˆ	ˆ	ˆ	ð ŒhŒlˆÝÐmÐm°C·N²NÑ4DÔ4D×4KÒ4KÑ4MÔ4MÐmÑmÔmÑmÔmð 	)Ý# D¤JÑ/Ô/×:Ò:Ñ<Ô<ˆHØŸšÑ(Ô(×.Ò.Ñ0Ô0ð )ð )‘��1Ý˜a¥¤Ñ.Ô.ð )µu´~ÀaÑ7HÔ7H×7LÒ7LÑ7NÔ7Nð )ÕSXÔSaÐbjÐklÔbmÑSnÔSn×SrÒSrÑStÔStð )Ø—G’G˜H QœKÑ(Ô(Ð(øõ ˜S‘>”>ˆÝ�s‰mŒm× Ò Ñ"Ô"×%Ò%µEÔ4KÐ%ÑLÔLˆÝ�3˜Ñ$Ô$ð 	!Ø ˆCŒMà—’Ñ!Ô!×(Ò(Ñ*Ô*ð 	%ð 	%ˆAÝ˜!�Uœ\Ñ*Ô*ð %¨q×/BÒ/BÑ/DÔ/Dð %ÝÔ! !Ñ$Ô$Ð$øõ 	�#‰ŒˆØ—’‘”ˆÝŒ
ðØ˜œðà Ô 1ðð ˜ðð �sð	ð
 ˜4œ8Ô+ðð ˜Hðð ÕAÅ(È4Ì>×KdÒKdÑKfÔKfÑBgÔBgÑhÔhðð ˜$œ+×0Ò0Ñ2Ô2ðð �d 4¤9™oœoðð  Ð!J D¤LÐ!J°)¸T¼\Ð!JÐ!Jðð   ×!6Ò!6Ñ!8Ô!8ðð �œ™œ×1Ò1Ñ3Ô3×=Ò=Ñ?Ô?ðð �;ðð Ý¥¤™MœMÝ!œjÝ!œjÝ"œ{Ý!œjðð ðð* ÐGð+ð, Ð6ð-ð0 ñ3	
ô 	
ð 	
ð6 !Ÿ/š/Ñ+Ô+ˆð 	Œ	�Š ¨tˆÑ4Ô4Ð4ØŒ	×Ò˜oÑ.Ô.Ð.ØÔ ¤Ò,Ð,ØŒI×!Ò! /Ñ2Ô2Ð2ØÔ˜qÒ Ð  t¤z°DÔ4DÑ'DÈÒ'IÐ'IØŒYÐ0 ¤Ð0Ð0Ð0Ñ0×=Ò=¸oÑNÔNÐNØˆtrR   c           	     óÖ  — 	 t          | j        j        | j        j        | j        j        ¬¦  «        | j        _        | j        j        dk    r&t          | j        j        | j        j        ¬¦  «        }n~t          | j        j        ¦  «                             dd¦  «        d         dv s| j        j        dv r;t          | j        j        | j        j        ¬¦  «        }d|v r|d         | j        _        nL# t          $ r?}t          t          d	t          | j        j        ¦  «        › d
|› �¦  «        ¦  «        |‚d}~ww xY w| j        j        r t          j        d¦  «         ddi|d<   d|d<   |S )z·Get train and validation datasets from data dictionary.

        Returns:
            (dict): A dictionary containing the training/validation/test dataset and category names.
        )r�   Úclassifyr«   rU   rY   >   Úymlr  >   rï   Úposerð   ÚdetectÚsegmentrñ   Ú	yaml_filez	Dataset 'u   ' error â�Œ Nz)Overriding class names with single class.r   Úitemr�  r€  )r   rb   rˆ   rŒ  r�   r  r   rs   Úrsplitr   rë  rå   r"   r    Ú
single_clsr   r´   )r˜   rˆ   rq  s      rP   r‡   zBaseTrainer.get_dataset4  sj  € ð	fÝ=¸d¼i¼nÈdÌiÔN`ÐhlÔhqÔhwÐxÑxÔxˆDŒIŒNð ŒyŒ~ Ò+Ð+Ý(¨¬¬¸t¼y¼ÐOÑOÔO��Ý�T”Y”^Ñ$Ô$×+Ò+¨C°Ñ3Ô3°BÔ7¸?ÐJÐJÈdÌiÌnð að Oð Oõ )¨¬¬¸t¼y¼ÐOÑOÔO�Ø $Ð&Ð&Ø%)¨+Ô%6�D”I”NøøÝð 	fð 	fð 	fÝ�vÐ&\µ)¸D¼I¼NÑ2KÔ2KÐ&\Ð&\ÐYZÐ&\Ð&\Ñ]Ô]Ñ^Ô^ÐdeÐeøøøøð	føøøàŒ9Ôð 	ÝŒKÐCÑDÔDÐDØ ˜KˆD�‰MØˆD�‰JØˆs   ‚C.C1 Ã1
D:Ã;:D5Ä5D:c                ó¸  — t          | j        t          j        j        ¦  «        rdS | j        d}}d}t          | j        ¦  «                             d¦  «        rt          | j        ¦  «        \  }}|j        }t          | j	        j
        t
          t          f¦  «        r$| j        st          | j	        j
        ¦  «        \  }}n| j	        j
        du r	| j        sd}t          |t          ¦  «        rÉt          dv rt          j        d¦  «         |                      ||j        t          dv ¬¦  «        }| j	        |_	        |j        �|j        n| j	        j        }t          ||¬¦  «        }t+          |dd¦  «        �1|j                             |j                             ¦   «         ¦  «         d|_        || _        n$|                      ||t          dv ¬¦  «        | _        |S )	z«Load, create, or download model for any task.

        Returns:
            (dict | None): Checkpoint to resume training from, or None if no checkpoint is loaded.
        Nr  FrX   z;Resuming training DistillationModel from checkpoint weights)r™   rW   Úverboser&  Ú	projector)re   r†   rß   r   ÚModulers   Úendswithr   r  rb   Ú
pretrainedr	   r”   r   r   r   r´   Ú	get_modelr'  r(  r8  rd   r'  Úload_state_dictr  r´  )r˜   r™   rW   rX  Ú_r'  r(  r†   s           rP   r2  zBaseTrainer.setup_modelS  sµ  € õ �d”j¥%¤(¤/Ñ2Ô2ð 	ØˆFà”z 4ˆWˆØˆÝˆtŒz‰?Œ?×#Ò# EÑ*Ô*ð 	Ý+¨D¬JÑ7Ô7‰MˆG�TØ”,ˆCÝ�d”iÔ*­Sµ$¨KÑ8Ô8ð 	ÀÄð 	Ý(¨¬Ô)=Ñ>Ô>‰JˆG�Q�QØŒYÔ! UÐ*Ð*°4´;Ð*ØˆGõ �gÕ0Ñ1Ô1ð 	[Ý�wˆˆÝ”ÐYÑZÔZÐZØ ŸNšN¨s¸GÔ<QÕ[_ÐcjÐ[j˜NÑkÔkˆMØ!%¤ˆMÔà5<Ô5JÐ5V˜GÔ1Ð1Ð\`Ô\eÔ\sˆMÝ%°MÐQ^Ð_Ñ_Ô_ˆEÝ�w ¨TÑ2Ô2Ð>Ø”×/Ò/°Ô0A×0LÒ0LÑ0NÔ0NÑOÔOÐOØ"ˆEŒOØˆDŒJˆJàŸš¨C¸Í$ÐRYÈ/˜ÑZÔZˆDŒJØˆrR   c                ó\  — | j                              | j        ¦  «         | j        j        dk    rFt
          r?t          j        j         	                    | j
                             ¦   «         dd¬¦  «         n=t          j        j         	                    | j
                             ¦   «         d¬¦  «         | j                              | j        ¦  «         | j                              ¦   «          | j                             ¦   «          | j        r!| j                             | j
        ¦  «         dS dS )zVPerform a single step of the training optimizer with gradient clipping and EMA update.rÐ   g      $@F)Úmax_normÚforeach)r/  N)rJ  Úunscale_r‹   rh   ri   r/   rß   r   ÚutilsÚclip_grad_norm_r†   rB  r–  r{  r�  r‰   rÆ   s    rP   r­  zBaseTrainer.optimizer_stepw  sû   € àŒ×Ò˜Tœ^Ñ,Ô,Ð,ØŒ;Ô˜uÒ$Ð$­Ð$ÝŒHŒN×*Ò*¨4¬:×+@Ò+@Ñ+BÔ+BÈTÐ[`Ð*ÑaÔaÐaÐaåŒHŒN×*Ò*¨4¬:×+@Ò+@Ñ+BÔ+BÈTÐ*ÑRÔRÐRØŒ×Ò˜œÑ(Ô(Ð(ØŒ×ÒÑÔÐØŒ× Ò Ñ"Ô"Ð"ØŒ8ð 	(ØŒH�OŠO˜DœJÑ'Ô'Ð'Ð'Ð'ð	(ð 	(rR   c                ó   — |S )zTAllow custom preprocessing of model inputs and ground truths depending on task type.rK   )r˜   ry   s     rP   r7  zBaseTrainer.preprocess_batch„  s   € àˆrR   c                ó¤  — | j         rB| j        dk    r7| j         j                              ¦   «         D ]}t          j        |d¬¦  «         Œ|                      | ¦  «        }|€dS |                     d| j                             ¦   «          	                    ¦   «          
                    ¦   «          ¦  «        }| j        �| j        |k     r|| _        ||fS )aP  Run validation on val set using self.validator.

        Returns:
            (tuple): A tuple containing:
                - metrics (dict | None): Dictionary of validation metrics, or None if validation was skipped.
                - fitness (float | None): Fitness score for the validation, or None if validation was skipped.
        rU   r   r#  Nr¬   rŽ   )r‰   rƒ   Úbuffersrä   rF  rk   Úpopr�   ÚdetachrS   Únumpyr�   )r˜   r  rl   rŽ   s       rP   r·  zBaseTrainer.validateˆ  sÑ   € ð Œ8ð 	.˜œ¨!Ò+Ð+àœ(œ,×.Ò.Ñ0Ô0ð .ð .�Ý”˜v¨1Ð-Ñ-Ô-Ð-Ð-Ø—.’. Ñ&Ô&ˆØˆ?Ø�:Ø—+’+˜i¨$¬)×*:Ò*:Ñ*<Ô*<×*@Ò*@Ñ*BÔ*B×*HÒ*HÑ*JÔ*JÐ)JÑKÔKˆØÔÐ$¨Ô(9¸GÒ(CÐ(CØ 'ˆDÔØ˜ÐÐrR   Tc                ó    — t          d¦  «        ‚)z>Get model and raise NotImplementedError for loading cfg files.z3This task trainer doesn't support loading cfg files©ÚNotImplementedError)r˜   r™   rW   r&  s       rP   r+  zBaseTrainer.get_modelœ  s   € å!Ð"WÑXÔXÐXrR   c                ó    — t          d¦  «        ‚)z>Raise NotImplementedError (must be implemented by subclasses).z1get_validator function not implemented in trainerr;  rÆ   s    rP   rN  zBaseTrainer.get_validator   ó   € å!Ð"UÑVÔVÐVrR   r  r»   c                ó    — t          d¦  «        ‚)zVRaise NotImplementedError (must return a `torch.utils.data.DataLoader` in subclasses).z2get_dataloader function not implemented in trainerr;  )r˜   Údataset_pathrz   rÝ   rî   s        rP   rû   zBaseTrainer.get_dataloader¤  s   € å!Ð"VÑWÔWÐWrR   c                ó    — t          d¦  «        ‚)zBuild dataset.z1build_dataset function not implemented in trainerr;  )r˜   Úimg_pathrî   ry   s       rP   Úbuild_datasetzBaseTrainer.build_dataset¨  r>  rR   c                ón   ‡— |€ˆfd„| j         D ¦   «         S ˆfd„|                     ¦   «         D ¦   «         S )zcReturn a loss dict with labeled training loss items, or a list of loss names if loss_items is None.Nc                ó   •— g | ]	}‰› d |› �‘Œ
S )rv  rK   )rM   rÀ   r/  s     €rP   rQ   z0BaseTrainer.label_loss_items.<locals>.<listcomp>¯  s#   ø€ Ð=Ð=Ð=¨�v�O�O �O�OÐ=Ð=Ð=rR   c                ó\   •— i | ](\  }}‰› d |› �t          t          |¦  «        d¦  «        “Œ)S )rv  é   )rË   rŸ  )rM   r›   rœ   r/  s      €rP   rn  z0BaseTrainer.label_loss_items.<locals>.<dictcomp>°  s9   ø€ ÐSÐSÐS¹¸¸1�6��˜A��¥¥u¨Q¡x¤x°Ñ!3Ô!3ÐSÐSÐSrR   )r‘   r   )r˜   r£  r/  s     `rP   rQ  zBaseTrainer.label_loss_items¬  sK   ø€ àÐØ=Ð=Ð=Ð=¨T¬_Ð=Ñ=Ô=Ð=ØSÐSÐSÐSÀ
×@PÒ@PÑ@RÔ@RÐSÑSÔSÐSrR   c                ó4   — | j         d         | j        _        dS )z/Set or update model parameters before training.r�  N)rˆ   r†   r�  rÆ   s    rP   r5  z BaseTrainer.set_model_attributes²  s   € àœ9 WÔ-ˆŒ
ÔÐÐrR   c                ó   — dS )zSCompute and set class weights for handling class imbalance. Override in subclasses.NrK   rÆ   s    rP   rO  zBaseTrainer.set_class_weights¶  ó   € € € rR   c                ó   — dS )z-Build target tensors for training YOLO model.NrK   )r˜   rÅ  Útargetss      rP   Úbuild_targetszBaseTrainer.build_targets¹  rJ  rR   c                ó   — dS )z-Return a string describing training progress.Ú rK   rÆ   s    rP   rœ  zBaseTrainer.progress_string¼  s   € àˆrrR   c                ó   — dS )z+Plot training samples during YOLO training.NrK   )r˜   ry   rÃ  s      rP   r³  z!BaseTrainer.plot_training_samplesÁ  rJ  rR   c                ó   — dS )z$Plot training labels for YOLO model.NrK   rÆ   s    rP   rS  z BaseTrainer.plot_training_labelsÄ  rJ  rR   c                ó„  — t          |                     ¦   «         ¦  «        t          |                     ¦   «         ¦  «        }}t          |¦  «        dz   }t	          j        ¦   «         | j        z
  }| j        j                             dd¬¦  «         | j         	                    ¦   «         rdn"d|z  ddg|¢R z   
                    d¦  «        d	z   }t          | j        d
d¬¦  «        5 }|                     |d|z  | j        dz   |g|¢R z   
                    d¦  «        z   d	z   ¦  «         ddd¦  «         dS # 1 swxY w Y   dS )z$Save training metrics to a CSV file.ra   TrZ   rO  z%s,r’  r‰  r_   r†  Úaúutf-8©Úencodingz%.6g,rU   N)r;  rP  r±  r€   r‰  r‹  r’   Úparentrr   r“   ÚrstripÚopenÚwriter’  )r˜   rl   rP  Úvalsrñ  rN   rp  rð  s           rP   r¹  zBaseTrainer.save_metricsÇ  s„  € å˜'Ÿ,š,™.œ.Ñ)Ô)­4°·²Ñ0@Ô0@Ñ+AÔ+AˆdˆÝ�‰LŒL˜1ÑˆÝŒI‰KŒK˜$Ô/Ñ/ˆØŒŒ×Ò d°TÐÑ:Ô:Ð:Ø”(—/’/Ñ#Ô#ÐbˆBˆB¨%°!©)°wÀÐ6NÈÐ6NÐ6NÑ*N×)VÒ)VÐWZÑ)[Ô)[Ð^bÑ)bˆÝ�$”(˜C¨'Ð2Ñ2Ô2ð 	W°aØ�GŠG�A˜ 1™¨¬
°Q©¸Ð'A¸DÐ'AÐ'AÑA×IÒIÈ#ÑNÔNÑNÐQUÑUÑVÔVÐVð	Wð 	Wð 	Wñ 	Wô 	Wð 	Wð 	Wð 	Wð 	Wð 	Wð 	Wð 	Wøøøð 	Wð 	Wð 	Wð 	Wð 	Wð 	Ws   Ã&AD5Ä5D9Ä<D9c                ó<   — t          | j        | j        ¬¦  «         dS )zPlot metrics from a CSV file.)rº   Úon_plotN)r-   r’   r]  rÆ   s    rP   r¼  zBaseTrainer.plot_metricsÑ  s   € å˜$œ(¨D¬LÐ9Ñ9Ô9Ð9Ð9Ð9rR   c                ó`   — t          |¦  «        }|t          j        ¦   «         dœ| j        |<   dS )z2Register plots (e.g. to be consumed in callbacks).)rˆ   Ú	timestampN)r	   r‰  rm   )r˜   rp   rˆ   Úpaths       rP   r]  zBaseTrainer.on_plotÕ  s-   € å�D‰zŒzˆØ$(µt´y±{´{ÐCÐCˆŒ
�4ÑÐÐrR   c                óÒ  — | j                              ¦   «         r| j         nd}t          t          ¦  «        5  t          dv r\| j                             ¦   «         rt          | j        ¦  «        ni }|r+t          | j         d|                     d¦  «        i¬¦  «         ddd¦  «         n# 1 swxY w Y   |r±t          j	        d|› d�¦  «         | j
        j        | j        j
        _        d| j        j
        _        |                      |¬¦  «        | _        | j                             d	d¦  «         | xj        d
z  c_        |                      d¦  «         | xj        d
z  c_        dS dS )z;Perform final evaluation and validation for the YOLO model.NrX   r   )rý  z
Validating z...F)r†   rŽ   rU   r…  )rw   r“   rA   r   r   rv   r?   r§   r   r´   rb   rm   rk   r9  rl   r7  r’  r…   )r˜   r†   rX  s      rP   r»  zBaseTrainer.final_evalÚ  s¦  € à!œY×-Ò-Ñ/Ô/Ð9�”	�	°TˆÝ)­*Ñ5Ô5ð 	eð 	eÝ�wˆˆØ59´Y×5EÒ5EÑ5GÔ5GÐO• t¤yÑ1Ô1Ð1ÈR�Øð eå# D¤I¸ÈÏÊÐRaÑIbÔIbÐ7cÐdÑdÔdÐdð	eð 	eð 	eñ 	eô 	eð 	eð 	eð 	eð 	eð 	eð 	eøøøð 	eð 	eð 	eð 	eð ð 	ÝŒKÐ2¨Ð2Ð2Ð2Ñ3Ô3Ð3Ø(,¬	¬ˆDŒNÔÔ%Ø*/ˆDŒNÔÔ'ØŸ>š>°˜>Ñ6Ô6ˆDŒLØŒL×Ò˜Y¨Ñ-Ô-Ð-ØˆJŒJ˜!‰OˆJŒJØ×ÒÐ1Ñ2Ô2Ð2ØˆJŒJ˜!‰OˆJŒJˆJˆJð	ð 	s   ·A&B)Â)B-Â0B-c                óú  — | j         j        }|�rd	 t          |t          t          f¦  «        o t	          |¦  «                             ¦   «         }t	          |rt          |¦  «        nt          ¦   «         ¦  «        }t          |¦  «        d         j         }n"# t          $ r}t          d¦  «        |‚d}~ww xY w| j         j        sBt          |d         t          ¦  «        s6t	          |d         ¦  «                             ¦   «         s| j         j        |d<   d}t          |¦  «        | _         t          |¦  «        x| j         _        | j         _        dD ]"}||v rt          | j         |||         ¦  «         Œ#|| _        dS )zCCheck if resume checkpoint exists and update arguments accordingly.r   zzResume checkpoint not found. Please pass a valid checkpoint to resume from, i.e. 'yolo train resume model=path/to/last.pt'Nrˆ   T)rþ   ry   rh   rŽ  rI   rx   r}   r  r0  r‰  r:  rò   rm   r4  r8  ro   )rb   r”   re   rs   r	   r“   r%   r+   r   rë  ÚFileNotFoundErrorrˆ   rf   r   r†   Úsetattr)r˜   rš   r”   r“   rv   Ú	ckpt_argsrq  r›   s           rP   rc   zBaseTrainer.check_resumeí  s}  € à”Ô!ˆØñ #	8ðÝ# F­Sµ$¨KÑ8Ô8ÐR½TÀ&¹\¼\×=PÒ=PÑ=RÔ=R�Ý°&ÐN�J vÑ.Ô.Ð.½nÑ>NÔ>NÑOÔO�Ý+¨DÑ1Ô1°!Ô4Ô9�	�	øÝð ð ð Ý'ðEñô ð ðøøøøðøøøð
 ŒyŒ~ð 3¥j°¸6Ô1BÅDÑ&IÔ&Ið 3ÕRVÐW`ÐagÔWhÑRiÔRi×RpÒRpÑRrÔRrð 3Ø$(¤I¤N�	˜&Ñ!àˆFÝ 	Ñ*Ô*ˆDŒIÝ14°T±´Ð:ˆDŒIŒO˜dœiÔ.ðð 8ð 8�ð$ ˜	�>�>Ý˜DœI q¨)°A¬,Ñ7Ô7Ð7øØˆŒˆˆs   ‘BB Â
B4ÂB/Â/B4c                ó¶  — |                      d¦  «        �dt          |d         d         | j        j        ¦  «        D ]\  }}|                      d¦  «        |d<   Œ| j                             |d         ¦  «         |                      d¦  «        r | j                             |d         ¦  «         | j        r‰|                      d¦  «        rtt          | j        ¦  «        | _        | j        j                             |d          	                    ¦   «          
                    ¦   «         ¦  «         |d         | j        _        |                      d¦  «        | _        dS )	z>Load optimizer, scaler, EMA, and best_fitness from checkpoint.r‹   Nr
  ÚfusedrJ  r‰   rý  r�   )r§   rR  r‹   r
  r,  rJ  r‰   r3   r†   rŸ  r  rý  r�   )r˜   rX  Úsavedr  s       rP   Ú_load_checkpoint_statez"BaseTrainer._load_checkpoint_state  s*  € à�8Š8�KÑ Ô Ð,Ý # D¨Ô$5°nÔ$EÀtÄ~ÔGbÑ cÔ cð 4ð 4‘��uØ!&§¢¨7Ñ!3Ô!3��g‘�ØŒN×*Ò*¨4°Ô+<Ñ=Ô=Ð=Ø�8Š8�HÑÔð 	8ØŒK×'Ò'¨¨X¬Ñ7Ô7Ð7ØŒ8ð 	/˜Ÿš ™œð 	/Ý ¤
Ñ+Ô+ˆDŒHð ŒHŒL×(Ò(¨¨e¬×):Ò):Ñ)<Ô)<×)GÒ)GÑ)IÔ)IÑJÔJÐJØ# IœˆDŒHÔØ ŸHšH ^Ñ4Ô4ˆÔÐÐrR   c                ó0  — | j         duo| j                              ¦   «          }| j        duot          j        | j        ¦  «         }t          dv o|p|}|rdnd}t          dk    r-t          dk    r|ndg}t          j        |d¦  «         |d         }|sdS || j        k    rt          j	        |› d�¦  «         dS | j
                             ¦   «         st          |› d	�¦  «        ‚| xj        d
z  c_        | j        dk    rt          d| j        › d�¦  «        ‚t          j	        |› d| j        › d�¦  «         |                      ¦   «          t          | j
        ¦  «        \  }}|d                              ¦   «         }	|	                     ¦   «         }
t%          d„ |
                     ¦   «         D ¦   «         ¦  «        st          d| j
        › d�¦  «        ‚t)          | j        ¦  «        }t-          |d¦  «        rc|j                             |	j                             ¦   «         ¦  «         |j                             |	j                             ¦   «         ¦  «         n|                     |
¦  «         |                      |¦  «         ~~	~
|d
z
  | j        _        dS )z@Detect and recover from NaN/Inf loss by loading last checkpoint.NrX   zLoss NaN/InfzFitness NaN/InfrY   r   Fz- detected but can not recover from last.pt...z8 detected but no valid last.pt is available for recoveryrU   rt  z#Training failed: NaN persisted for z epochsz detected (attempt z/3), recovering from last.pt...r‰   c              3  ó–   K  — | ]D}t          |t          j        ¦  «        ¯t          j        |¦  «                             ¦   «         V — ŒEd S r  rù  r  s     rP   r  z3BaseTrainer._handle_nan_recovery.<locals>.<genexpr>?  sL   è è € ÐfÐf¨qÍ*ÐUVÕX]ÔXdÑJeÔJeÐf•5”> !Ñ$Ô$×(Ò(Ñ*Ô*ÐfÐfÐfÐfÐfÐfrR   zCheckpoint z" is corrupted with NaN/Inf weightsr'  T)r�   rû  rŽ   r�  r   rä   r®  r|   r   r²   rv   r“   rå   r—   r—  r   rŸ  r  rü  r±  rC   r†   rã   r'  r,  r'  ri  rŒ   rW  )r˜   r’  Úloss_nanÚfitness_nanÚ	corruptedÚreasonrÉ  r-  rX  r‰   Ú	ema_stater†   s               rP   r¸  z BaseTrainer._handle_nan_recovery&  s­  € à”9 DÐ(ÐE°´×1CÒ1CÑ1EÔ1EÐ-EˆØ”l¨$Ð.ÐPµr´{À4Ä<Ñ7PÔ7PÐ3PˆÝ˜G�OÐA¨Ð)@°[ˆ	Ø#+ÐB��Ð1BˆÝ�2Š:ˆ:Ý+/°1ª9¨9˜i˜i¸$Ð?ˆNÝÔ& ~°qÑ9Ô9Ð9Ø& qÔ)ˆIØð 	Ø�5Ø�DÔ$Ò$Ð$ÝŒN˜fÐSÐSÐSÑTÔTÐTØ�5ØŒy×ÒÑ!Ô!ð 	dÝ &ÐbÐbÐbÑcÔcÐcØÐ"Ô" aÑ'Ð"Ô"ØÔ%¨Ò)Ð)ÝÐhÀTÔE_ÐhÐhÐhÑiÔiÐiÝŒ˜&ÐpÐp°TÔ5OÐpÐpÐpÑqÔqÐqØ×ÒÑÔÐÝ! $¤)Ñ,Ô,‰ˆˆ4Ø�5Œk×ÒÑ!Ô!ˆØ—N’NÑ$Ô$ˆ	ÝÐfÐf°I×4DÒ4DÑ4FÔ4FÐfÑfÔfÑfÔfð 	\ÝÐZ¨T¬YÐZÐZÐZÑ[Ô[Ð[Ý˜TœZÑ(Ô(ˆÝ�5˜/Ñ*Ô*ð 	-ð Ô×/Ò/°Ô0A×0LÒ0LÑ0NÔ0NÑOÔOÐOØŒO×+Ò+¨C¬M×,DÒ,DÑ,FÔ,FÑGÔGÐGÐGà×!Ò! )Ñ,Ô,Ð,Ø×#Ò# DÑ)Ô)Ð)Ø�#�yØ$)¨A¡IˆŒÔ!ØˆtrR   c           	     óf  — |�| j         sdS |                     dd¦  «        dz   }d|cxk     r| j        k     s.n J | j        j        › d| j        › d| j        j        › d�¦   «         ‚t          j        d	| j        j        › d
|dz   › d| j        › d�¦  «         |                      |¦  «         t          | j        ¦  «        }t          t          |d|¦  «        j        d         dd¦  «        �A| 
                    ¦   «         |_        |dz
  |j        _        |j                             ¦   «          || _        dS )z-Resume YOLO training from a given checkpoint.Nr’  rY   rU   r   z training to zf epochs is finished, nothing to resume.
Start a new training without resuming, i.e. 'yolo train model=r  zResuming training z from epoch z to z total epochsr'  Úone2one_cv2)r”   r§   r{   rb   r†   r   r´   ri  rC   rd   Úinit_criterionr´  rý  r{  r|   )r˜   rX  r|   r†   s       rP   rV  zBaseTrainer.resume_trainingN  s]  € àˆ<˜tœ{ˆ<ØˆFØ—h’h˜w¨Ñ+Ô+¨aÑ/ˆØ�;Ð,Ð,Ò,Ð, ¤Ò,Ð,Ð,Ð,Ð,ØŒyŒð `ð `¨T¬[ð `ð `ØMQÌYÌ_ð`ð `ð `ñ -Ô,Ð,õ 	ŒÐu¨¬¬ÐuÐuÀkÐTUÁoÐuÐuÐ[_Ô[fÐuÐuÐuÑvÔvÐvØ×#Ò# DÑ)Ô)Ð)Ý˜TœZÑ(Ô(ˆÝ•7˜5 /°5Ñ9Ô9Ô?ÀÔCÀ]ÐTXÑYÔYÐeà#×2Ò2Ñ4Ô4ˆEŒOØ&1°A¡oˆEŒOÔ#ØŒO×"Ò"Ñ$Ô$Ð$Ø&ˆÔÐÐrR   c                ó   — t          | j        j        d¦  «        rd| j        j        _        t          | j        j        d¦  «        rHt	          j        d¦  «         | j        j                             t          | j        ¦  «        ¬¦  «         dS dS )z5Update dataloaders to stop using mosaic augmentation.ÚmosaicFrŽ  zClosing dataloader mosaic)ÚhypN)	rã   rü   r  ru  r   r´   rŽ  r   rb   rÆ   s    rP   rš  z$BaseTrainer._close_dataloader_mosaica  s‰   € å�4Ô$Ô,¨hÑ7Ô7ð 	5Ø/4ˆDÔÔ%Ô,Ý�4Ô$Ô,¨nÑ=Ô=ð 	HÝŒKÐ3Ñ4Ô4Ð4ØÔÔ%×2Ò2µt¸D¼I±´Ð2ÑGÔGÐGÐGÐGð	Hð 	HrR   Úautoçü©ñÒMbP?çÍÌÌÌÌÌì?çñhãˆµøä>ç     jø@c                ó	  — i i i i g}t          d„ t          j                             ¦   «         D ¦   «         ¦  «        }h d£}	d„ |	D ¦   «                              t          |¦  «                             ¦   «         t          |¦  «        ¦  «        }|dk    rŒt          j        t          d¦  «        › d| j
        j        › d| j
        j        › d�¦  «         | j                             d	d
¦  «        }
t          dd|
z   z  d¦  «        }|dk    rdnd|df\  }}}d| j
        _        |dk    }t!          |¦  «                             ¦   «         D ]~\  }}|                     d¬¦  «        D ]b\  }}|r|› d|› �n|}|j        dv r|r||d         |<   Œ'd|v r||d         |<   Œ7t)          ||¦  «        sd|v r||d         |<   ŒW||d         |<   ŒcŒ|sd„ |dd…         D ¦   «         }|d v r	||d!fdd"œ}n5|d#k    r||d$œ}n)|d%k    s|dk    r||d&d'œ}nt+          d(|› d)|	› d*�¦  «        ‚t-          |d         ¦  «        t-          |d         ¦  «        t-          |d         ¦  «        g}d+|d         i|¥d,di¥|d<   d+|d         i|¥|d-d.œ¥|d<   d+|d         i|¥dd/d.œ¥|d<   d0\  }}|�rt-          |d         ¦  «        |d<   d+|d         i|¥|d&d1d2œ¥|d<   t!          |¦  «        }t/          |d3|¦  «        j        d4         }t/          |d5d¦  «        t/          |d6d¦  «        f}d7„ |D ¦   «         }g }|D ]‹}|                     d+¦  «        }g g }}|                     ¦   «         D ]5\  }} t5          | ¦  «        |v sd8|v sd9|v r|n|                     | ¦  «         Œ6|                     d+|i|¥d:|dz  i¥d+|i|¥g¦  «         ŒŒ|}|d;v r1t:          r*t=          |                     ¦   «         ¦  «        j         rd<d&ini }! |rtC          tD          ||¬=¦  «        nt/          tF          |¦  «        dFd+|i|!¤Ž}"t          j        t          d¦  «        › d>tI          |"¦  «        j%        › d?|› d@|› dA|d         › dB|d         › dC|› dD|d         › dE�¦  «         |"S )GaÖ  Construct an optimizer for the given model.

        Args:
            model (torch.nn.Module): The model for which to build an optimizer.
            name (str, optional): The name of the optimizer to use. If 'auto', the optimizer is selected based on the
                number of iterations.
            lr (float, optional): The learning rate for the optimizer.
            momentum (float, optional): The momentum factor for the optimizer.
            decay (float, optional): The weight decay for the optimizer.
            iterations (float, optional): The number of iterations, which determines the optimizer if name is 'auto'.

        Returns:
            (torch.optim.Optimizer): The constructed optimizer.
        c              3  ó(   K  — | ]\  }}d |v ¯	|V — ŒdS )ÚNormNrK   ©rM   r›   rœ   s      rP   r  z.BaseTrainer.build_optimizer.<locals>.<genexpr>y  s*   è è € ÐDÐD™˜˜A¸À!¸¸�1¸¸¸¸ÐDÐDrR   >	   ÚSGDÚAdamrw  ÚAdamWr   ÚNAdamÚRAdamÚAdamaxÚRMSpropc                ó8   — i | ]}|                      ¦   «         |“ŒS rK   )r©  r  s     rP   rn  z/BaseTrainer.build_optimizer.<locals>.<dictcomp>{  s"   € Ð1Ð1Ð1 �—’‘	”	˜1Ð1Ð1Ð1rR   rw  z
optimizer:z' 'optimizer=auto' found, ignoring 'lr0=z' and 'momentum=zJ' and determining best 'optimizer', 'lr0' and 'momentum' automatically... r€  é
   ç{®Gáz„?é   é   i'  )r   r‰  ry  r‚  ry  ri  r   F)Úrecurser«   >   ra   rŠ  rt  rh  ra   Úlogit_scalerU   r   c                ó6   — g | ]}|                      ¦   «         ‘ŒS rK   )r±  r  s     rP   rQ   z/BaseTrainer.build_optimizer.<locals>.<listcomp>•  s    € Ð+Ð+Ð+ �—’‘”Ð+Ð+Ð+rR   N>   r�  r‚  rƒ  r„  r…  g+‡ÙÎ÷ï?)ró   Úbetasrú   r†  )ró   rô   r€  T)ró   rô   ÚnesterovzOptimizer 'z,' not found in list of available optimizers zZ. Request support for additional optimizers at https://github.com/ultralytics/ultralytics.Úparamsr÷   rù   )rú   r÷   Úbn)gš™™™™™É?r©   rø   )rú   Úuse_muonr÷   r'  rY   Úcv3Úone2one_cv3c                ó^   — h | ]*}|¯|                      ¦   «         D ]}t          |¦  «        ’ŒŒ+S rK   )rB  Úid)rM   rö  Úps      rP   ú	<setcomp>z.BaseTrainer.build_optimizer.<locals>.<setcomp>¯  s8   € ÐIÐIÐI ¨qÐI¸!¿,º,¹.¼.ÐIÐI°Q•r˜!‘u”uÐIÐIÐIÐIrR   zproto.semsegÚSemanticSegmentró   >   r�  r‚  rg  )rø   Úsgdr­   z(lr=z, momentum=z) with parameter groups z weight(decay=0.0), z weight(decay=z), z bias(decay=0.0)rK   )&r¥  r   Ú__dict__r   r§   rs   r©  r   r´   r!   rb   r	  rô   rˆ   rË   r   rC   rò  r>  Úndimre   r<  r€   rd   r†   r7  r—  r    r�  r/   ÚnextrB  Úis_cudar   r   r   ri   Ú__name__)#r˜   r†   rp   ró   rô   rõ   rö   Úgr’  Ú
optimizersr€  Úlr_fitr“  Úmodule_nameÚmoduleÚ
param_nameÚparamÚfullnameÚ
optim_argsÚ
num_paramsrø   r›  ÚtargetÚheadÚheadsÚboostedÚg_rÀ   r˜  Úp1Úp2r›   rœ   rg  r‹   s#                                      rP   r  zBaseTrainer.build_optimizeri  s.  € ð ��R˜ÐˆÝÐDÐD¥¤×!2Ò!2Ñ!4Ô!4ÐDÑDÔDÑDÔDˆØeÐeÐeˆ
Ø1Ð1 jÐ1Ñ1Ô1×5Ò5µc¸$±i´i·o²oÑ6GÔ6GÍÈTÉÌÑSÔSˆØ�6Š>ˆ>ÝŒKÝ˜LÑ)Ô)ð Xð XØ!%¤¤ðXð XØ@DÄ	Ô@RðXð Xð Xñô ð ð
 ”—’˜t RÑ(Ô(ˆBÝ˜9¨¨B©Ñ/°Ñ3Ô3ˆFØ9CÀeÒ9KÐ9KÐ!5Ð!5ÐRYÐ[aÐcfÐQgÑˆD�"�hØ'*ˆDŒIÔ$à˜7’?ˆÝ#/°Ñ#6Ô#6×#DÒ#DÑ#FÔ#Fð 	+ð 	+ÑˆK˜Ø%+×%<Ò%<ÀUÐ%<Ñ%KÔ%Kð 
+ð 
+Ñ!�
˜EØ<GÐW˜kÐ8Ð8¨JÐ8Ð8Ð8ÈZ�Ø”: Ð'Ð'¨HÐ'Ø%*�A�a”D˜‘N�NØ˜xÐ'Ð'Ø%*�A�a”D˜‘N�NÝ ¨Ñ+Ô+ð +¨}ÀÐ/HÐ/Hà%*�A�a”D˜‘N�Nà%*�A�a”D˜‘N�Nð
+ð ð 	,Ø+Ð+ Q r¨ r¤UÐ+Ñ+Ô+ˆAàÐ@Ð@Ð@Ø "¨h¸Ð->ÐPSÐTÐTˆJˆJØ�YÒÐØ "°Ð9Ð9ˆJˆJØ�UŠ]ˆ]˜d gšo˜oØ "°ÀdÐKÐKˆJˆJå%ðk˜dð kð kÐPZð kð kð kñô ð õ
 ˜!˜Aœ$‘i”i¥ Q q¤T¡¤­C°°!´©I¬IÐ6ˆ
Ø˜!˜Aœ$ÐD *ÐD¨m¸VÐDÐDˆˆ!‰Ø˜!˜Aœ$Ð] *Ð]¸eÐT\Ð]Ð]Ð]ˆˆ!‰Ø˜!˜Aœ$ÐW *ÐW¸cÐRVÐWÐWÐWˆˆ!‰Ø‰	ˆˆcØñ 	Ý  !¤™IœIˆJ�q‰MØ˜a œdÐq jÐqÀ%ÐUYÐjpÐqÐqÐqˆAˆa‰Då! %Ñ(Ô(ˆFÝ˜6 ?°FÑ;Ô;ÔAÀ"ÔEˆDÝ˜T 5¨$Ñ/Ô/µ¸¸}ÈdÑ1SÔ1SÐTˆEØIÐI eÐIÑIÔIˆGØˆBØð Tð T�Ø—E’E˜(‘O”O�Ø˜R�B�ØŸGšG™IœIð pð p‘D�A�qÝ˜a™5œ5 GÐ+Ð+¨~ÀÐ/BÐ/BÐFWÐ[\ÐF\ÐF\�R�RÐbd×lÒlÐmnÑoÔoÐoÐoØ—	’	˜H bÐ<¨AÐ<¨t°R¸!±VÐ<Ð<¸xÈÐ>QÈqÐ>QÐRÑSÔSÐSÐSØˆAð $(Ð+<Ð#<Ð#<ÅÐ#<ÍtÐTY×TdÒTdÑTfÔTfÑOgÔOgÔOoÐ#<�˜$��ÐuwˆØ]¸8Ð]•W�U¨°3Ð7Ñ7Ô7Ð7ÍÕQVÐX\ÑI]ÔI]ÐqÐqÐfgÐqÐkpÐqÐqˆ	åŒÝ˜Ñ%Ô%ð yð y­¨Y©¬Ô(@ð yð yÀbð yð yÐU]ð yð yØ˜!Œ}ðyð yØ2<¸Q´-ðyð yØOTðyð yØYcÐdeÔYfðyð yð yñ	
ô 	
ð 	
ð ÐrR   ©rF   rG   )rž   rs   rÔ  )Fr  )rß  rà  )NNT)r  r   r»   )r»   N)Nr»   )rw  rx  ry  rz  r{  )0r   Ú
__module__Ú__qualname__Ú__doc__r   r�   r£   r¥   r…   r»   rÇ   rÍ   rë   r  r`  r¸   rM  r°  r«  rí  r—  rº  r‡   r2  r­  r7  r·  r+  rN  rû   rC  rQ  r5  rO  rM  rœ  r³  rS  r¹  r¼  r]  r»  rc   ri  r¸  rV  rš  r  rK   rR   rP   rE   rE   O   s@  € € € € € ð0ð 0ðd '°$ÐRVð S'ð S'ð S'ð S'ð S'ðj/ð /ð /ð /ð+ð +ð +ð +ðð ð ð ð
+ð +ð +ð@Xð Xð XðNð Nð Nð

ð 
ð 
ð(#9ð #9ð #9ðJ}6ð }6ð }6ð~nð nð nð`
ð 
ð 
ð 
ðXð Xð Xð Xð+ð +ð +ð +ð +ðð ð ðð ð ð@ð @ð @ðDð ð ð>"ð "ð "ðH(ð (ð (ðð ð ð ð  ð  ð(Yð Yð Yð YðWð Wð WðXð Xð Xð XðWð Wð Wð WðTð Tð Tð Tð.ð .ð .ðbð bð bð<ð <ð <ðð ð ð
:ð :ð :ð3ð 3ð 3ðWð Wð Wð:ð :ð :ðDð Dð Dð Dð
ð ð ð&'ð 'ð 'ðR5ð 5ð 5ð &ð &ð &ðP'ð 'ð 'ð&Hð Hð HðXð Xð Xð Xð Xð XrR   rE   c                  ó.   — e Zd ZdZd	d
d„Zd„ Zd„ Zd„ ZdS )ÚMultiTrainera˜  Fine-tune a single base model across a collection of datasets and aggregate per-dataset results.

    Used automatically by Model.train() when `data` is a list or tuple, allowing one base model to be benchmarked across
    many datasets (such as the RF100 collection) in a single call. The datasets are fine-tuned in series and the same
    base weights seed each run, so every run starts from an identical model. All output is grouped under one sweep
    directory (e.g. runs/detect/multitrain): each dataset gets its own run subdirectory, and the per-dataset and mean
    metrics are written to multitrain_results.json (for post-processing) alongside a multitrain_results.png bar
    chart. The base model object is left unchanged; each dataset's fine-tuned weights live in its own run directory.

    Attributes:
        trainer (type[BaseTrainer] | None): Task trainer class for Python runs, or None for CLI subprocess runs.
        args (dict): Training arguments shared across datasets; its `data` key holds the dataset collection.
        model (torch.nn.Module): Base model whose weights seed each per-dataset fine-tune.
        callbacks (dict | None): Callbacks forwarded to each per-dataset trainer.
        trainers (list[SimpleNamespace]): Completed per-dataset run records.
        metrics (dict): Mapping of each run name (e.g. coco8, coco8-2) to its training-metrics dict from the checkpoint.
        mean_metrics (dict): Mean training metrics across successful datasets.
        save_dir (Path | None): Sweep directory holding the per-dataset runs and the results JSON/plot.

    Examples:
        Fine-tune one base model across several datasets and read back per-run metrics:
        >>> from ultralytics import YOLO
        >>> model = YOLO("yolo26n.pt")
        >>> results = model.train(data=["coco8.yaml", "african-wildlife.yaml"], epochs=10)
        >>> results["coco8"]["fitness"]  # final fitness on the coco8 run
    NrF   rG   c                óv   — || _         || _        || _        || _        g | _        i | _        i | _        d| _        dS )a!  Initialize MultiTrainer with a task trainer class, shared training arguments, and the base model.

        Args:
            trainer (type[BaseTrainer] | None): Task trainer class to run once per dataset. None uses CLI subprocesses.
            args (dict): Training arguments; the `data` key holds the list/tuple of datasets to fine-tune on.
            model (torch.nn.Module): Base model whose weights seed each per-dataset fine-tune.
            _callbacks (dict, optional): Callback functions forwarded to each per-dataset trainer.
        N)Útrainerrb   r†   r   Útrainersrl   Úmean_metricsro   )r˜   r¹  rb   r†   rF   s        rP   r�   zMultiTrainer.__init__à  sA   € ð ˆŒØˆŒ	ØˆŒ
Ø#ˆŒØˆŒØˆŒØˆÔØˆŒˆˆrR   c                ó  — ddl m} ddlm}m} | j        d         } || j                             d¦  «        | j                             d¦  «        d| j                             dd	¦  «        ¬
¦  «        }t          |d¬¦  «        | _        | j         	                    dd¬¦  «         t          t          | j                             d¦  «        pd¦  «        ¦  «        j        }| j        €| j        |› d�z  nd}|rmt          | j        ¦  «        }t!          | j        ¦  «                             ¦   «         }	t%          |	¦  «          ||	|t'          | j        di ¦  «        dœ|¦  «         	 t)          |¦  «        D �]a\  }
}t+          j        dt/          ddd|
dz   › dt1          |¦  «        › d�¦  «        › d|› �¦  «         t          t          |¦  «        ¦  «        }|j        j        }|j        dk    r8|r6t          t6          j                             |j        ¦  «        ¦  «        j        n|j        }|}	 i | j        ¥|t          | j        ¦  «        |d	dœ¥} ||d         |d         |                     d¦  «        d|                     dd	¦  «        d¬¦  «        }t          |¦  «        }| 	                    dd¬¦  «         |j        }t          |¦  «        |d <   | j        €ft          |¦  «        |d<   |                     d!d¦  «         t?          j         g tB          ¢d‘d"„ | "                    ¦   «         D ¦   «         ¢d¬#¦  «         n§|                      || j#        ¬$¦  «        }|                     d%d¦  «        }|d	u rdn| j        }tI          |t          t          f¦  «        rtK          |¦  «        \  }}| &                    || j        j'        ¬&¦  «        |_        | (                    ¦   «          |d'z  d(z  |d'z  d)z  }}| )                    ¦   «         r|n|}d}| j        �)t'          t'          |d*d¦  «        d+d¦  «        }|�|j*        }|p&| )                    ¦   «         r ||¦  «        d,         nd| j+        |<   | j,         -                     ||||¬-¦  «        ¦  «         �Œ'# t\          $ r/}t+          j/        d.|› d/|› �¦  «         d| j+        |<   Y d}~�Œ[d}~ww xY w	 |r| 0                    d¬0¦  «         n# |r| 0                    d¬0¦  «         w w xY wtb          d1v rf| j,        r_| j         	                    dd¬¦  «         |  2                    ¦   «          | j                             d2d¦  «        r|  3                    ¦   «          | j+        S )3z[Fine-tune the base model on each dataset in series and return a {dataset: metrics} mapping.r   )ÚSimpleNamespace)Ú
torch_loadÚ
torch_saverˆ   Úprojectr  r»   r\   F)rÀ  r  rî   r\   Ú
multitrain)rp   TrZ   r†   Úmultitrain_baseNr  rb   )r†   r  rþ  r†  Úbluerc  zMultiTrainer rU   rv  rÏ   z fine-tuning on )rˆ   rÀ  rp   r”   rp   )rÀ  rp   r  rî   r\   ro   ro   r™   c              3  ó*   K  — | ]\  }}|› d |› �V — ŒdS )ú=NrK   r  s      rP   r  z%MultiTrainer.train.<locals>.<genexpr>-  s0   è è € Ð!KÐ!K±°°A Q * *¨ * *Ð!KÐ!KÐ!KÐ!KÐ!KÐ!KrR   r®   )rš   rF   r*  )rW   r™   rW   r^   r]   rk   rl   rÿ  )ro   rw   rv   zMultiTrainer: fine-tuning on z failed, skipping: )Ú
missing_okrX   rm   )4Útypesr½  Úultralytics.utils.patchesr¾  r¿  rb   r§   r   ro   rr   r	   rs   Ústemr¹  r<   r†   r   r  r@   rd   r™  r   r´   r!   r€   rW  rp   rá   r`  Úabspathr7  r¶   r·   r   r   r   re   r   r+  r  r»   r“   Úresults_dictrl   rº  r    rë  rÈ  r•   r   Úsave_resultsr-   )r˜   r½  r¾  r¿  ÚdatasetsÚsweepÚ
model_nameÚ
base_modelÚstater†   rm  rˆ   r`  rW  rp   Úrun_namerš   r·   ro   r¹  r*  rW   r-  rw   rv   rX  rl   rq  s                               rP   r»   zMultiTrainer.trainò  s  € à)Ð)Ð)Ð)Ð)Ð)àDÐDÐDÐDÐDÐDÐDÐDà”9˜VÔ$ˆà�Ø”I—M’M )Ñ,Ô,Ø”—’˜vÑ&Ô&ØØ”Y—]’] :¨uÑ5Ô5ð	
ñ 
ô 
ˆõ % U°Ð>Ñ>Ô>ˆŒØŒ×Ò D°4ÐÑ8Ô8Ð8Ý�#˜dœiŸmšm¨GÑ4Ô4ÐIÐ8IÑJÔJÑKÔKÔPˆ
Ø;?¼<Ð;O�T”]¨
Ð%7Ð%7Ð%7Ñ7Ð7ÐUYˆ
Øð 	wÝ˜dœjÑ)Ô)ˆEÝ˜TœZÑ(Ô(×-Ò-Ñ/Ô/ˆEÝ�eÑÔÐØˆJ °EÍÐQUÔQ[Ð]cÐegÑIhÔIhÐiÐiÐkuÑvÔvÐvð>	3Ý$ XÑ.Ô.ð :2ñ :2‘��4Ý”Øs� &¨&Ð2ZÀ!ÀaÁ%Ð2ZÐ2ZÍ#ÈhÉ-Ì-Ð2ZÐ2ZÐ2ZÑ[Ô[ÐsÐsÐmqÐsÐsñô ð õ �C ™IœI‘”�ØœÔ)�ØBFÄ)ÈvÒBUÐBUÐZ`ÐBU•t�BœGŸOšO¨D¬KÑ8Ô8Ñ9Ô9Ô>Ð>ÐfjÔfo�Ø�ð22ð!Øœ)ð!à $Ý#& t¤}Ñ#5Ô#5Ø $Ø"'ð!ð !ð !�Ið *˜/Ø )¨)Ô 4Ø& vÔ.Ø&Ÿ]š]¨6Ñ2Ô2Ø$Ø!*§¢¨z¸5Ñ!AÔ!AØ!%ðñ ô �Cõ  ,¨CÑ0Ô0�HØ—N’N¨4¸$�NÑ?Ô?Ð?Ø'œ}�HÝ,/°©M¬M�I˜jÑ)Ø”|Ð+Ý-0°©_¬_˜	 'Ñ*Ø!Ÿš e¨TÑ2Ô2Ð2Ý"œðÝ!2ðà 'ðð "LÐ!K¸¿ºÑ9JÔ9JÐ!KÑ!KÔ!Kðð
 #'ðñ ô ð ð ð #'§,¢,¸ÈtÌ~ ,Ñ"^Ô"^˜Ø%.§]¢]°<ÀÑ%FÔ%F˜
Ø*4¸Ð*=Ð*= $ $À4Ä:˜Ý% jµ3½°+Ñ>Ô>ð EÝ)8¸Ñ)DÔ)D™J˜G QØ(/×(9Ò(9À'ÈtÌzÌÐ(9Ñ(_Ô(_˜œØŸš™œ˜Ø!)¨IÑ!5¸	Ñ!AÀ8ÈiÑCWÐZcÑCc˜$�DØ#'§;¢;¡=¤=Ð:˜4˜4°d�DØ"�GØ”|Ð/Ý")­'°'¸;ÈÑ*MÔ*MÈyÐZ^Ñ"_Ô"_˜Ø"Ð.Ø&-Ô&:˜GØ-4Ð-vÐ^b×^iÒ^iÑ^kÔ^kÐ9u¸¸ÀDÑ9IÔ9IÈ/Ô9ZÐ9ZÐqu�D”L Ñ*Ø”M×(Ò(¨¨À(ÐQUÐ\`Ð)aÑ)aÔ)aÑbÔbÐbÑbøÝ ð 2ð 2ð 2Ý”LÐ!]ÀÐ!]Ð!]ÐZ[Ð!]Ð!]Ñ^Ô^Ð^Ø-1�D”L Ñ*Ð*Ð*Ð*Ð*Ñ*øøøøð2øøøðq:2ðx ð 3Ø×!Ò!¨TÐ!Ñ2Ô2Ð2øøð ð 3Ø×!Ò!¨TÐ!Ñ2Ô2Ð2Ð2ð3øøøå�7ˆ?ˆ?˜tœ}ˆ?ØŒM×Ò¨°tÐÑ<Ô<Ð<Ø×ÒÑÔÐØŒy�}Š}˜W dÑ+Ô+ð $Ø×!Ò!Ñ#Ô#Ð#ØŒ|Ðs8   Å'C	S3 È1I*RÒS3 Ò
SÒ'$SÓS3 ÓSÓS3 Ó3Tc                óÂ  ‡— ddl }d„ | j                             ¦   «         D ¦   «         }d„ |                     ¦   «         D ¦   «         Šd„ ‰D ¦   «         }ˆfd„|D ¦   «         | _        | j        dz  }t          |dd	¬
¦  «        5 }|                     || j        dœ|d¬¦  «         ddd¦  «         n# 1 swxY w Y   t          j	        dt          d|¦  «        › �¦  «         |S )z_Write per-dataset and mean metrics to multitrain_results.json for programmatic post-processing.r   Nc                óZ   — i | ](\  }}||rd „ |                      ¦   «         D ¦   «         nd“Œ)S )c                ó4   — i | ]\  }}|t          |¦  «        “ŒS rK   )rŸ  r  s      rP   rn  z8MultiTrainer.save_results.<locals>.<dictcomp>.<dictcomp>S  s$   € Ð<Ð<Ð<©$¨!¨Q˜!�U 1™XœXÐ<Ð<Ð<rR   N)r   )rM   r·   rö  s      rP   rn  z-MultiTrainer.save_results.<locals>.<dictcomp>S  sB   € ÐpÐpÐpÑQWÐQTÐVW�3ÀÐKÐ<Ð<°!·'²'±)´)Ð<Ñ<Ô<Ð<ÀtÐpÐpÐprR   c                ó   — g | ]}|¯|‘ŒS rK   rK   )rM   rö  s     rP   rQ   z-MultiTrainer.save_results.<locals>.<listcomp>T  s   € Ð2Ð2Ð2�q°Ð2�Ð2Ð2Ð2rR   c                ó   — h | ]	}|D ]}|’ŒŒ
S rK   rK   ©rM   rö  r›   s      rP   r™  z,MultiTrainer.save_results.<locals>.<setcomp>U  s%   € Ð,Ð,Ð,�a¨!Ð,Ð, Q�Ð,Ð,Ð,Ð,rR   c                ó‚   •‡— i | ]:Š‰t          ˆfd „‰D ¦   «         ¦  «        t          ˆfd„‰D ¦   «         ¦  «        z  “Œ;S )c              3  ó0   •K  — | ]}‰|v ¯|‰         V — Œd S r  rK   rØ  s     €rP   r  z7MultiTrainer.save_results.<locals>.<dictcomp>.<genexpr>V  s+   øè è € Ð#BÐ#B¨Q¸1À¸6¸6 A a¤D¸6¸6¸6¸6Ð#BÐ#BrR   c              3  ó    •K  — | ]}‰|v V — Œ	d S r  rK   rØ  s     €rP   r  z7MultiTrainer.save_results.<locals>.<dictcomp>.<genexpr>V  s(   øè è € ÐH_ÐH_ÐTUÈÈaÈÐH_ÐH_ÐH_ÐH_ÐH_ÐH_rR   )r¤  )rM   r›   Úvalids    @€rP   rn  z-MultiTrainer.save_results.<locals>.<dictcomp>V  s]   øø€ ÐnÐnÐnÐde˜Q¥Ð#BÐ#BÐ#BÐ#B°%Ð#BÑ#BÔ#BÑ BÔ BÅSÐH_ÐH_ÐH_ÐH_ÐY^ÐH_ÑH_ÔH_ÑE_ÔE_Ñ _ÐnÐnÐnrR   zmultitrain_results.jsonÚwrT  rU  )ÚresultsÚmeanra   )ÚindentúMultiTrainer results saved to rc  )Újsonrl   r   r±  r»  ro   rY  Údumpr   r´   r!   )r˜   râ  rÞ  rP  rº   rð  rÜ  s         @rP   rÌ  zMultiTrainer.save_resultsO  sL  ø€ àˆˆˆàpÐpÐ[_Ô[g×[mÒ[mÑ[oÔ[oÐpÑpÔpˆØ2Ð2˜GŸNšNÑ,Ô,Ð2Ñ2Ô2ˆØ,Ð,˜5Ð,Ñ,Ô,ˆØnÐnÐnÐnÐimÐnÑnÔnˆÔØŒ}Ð8Ñ8ˆÝ�$˜ gÐ.Ñ.Ô.ð 	T°!Ø�IŠI '°4Ô3DÐEÐEÀqÐQRˆIÑSÔSÐSð	Tð 	Tð 	Tñ 	Tô 	Tð 	Tð 	Tð 	Tð 	Tð 	Tð 	Tøøøð 	Tð 	Tð 	Tð 	TåŒÐMµX¸fÀdÑ5KÔ5KÐMÐMÑNÔNÐNØˆs   Â!B/Â/B3Â6B3c                ó@  ‡— ddl m} ddlm} |                     | j                             d¦  «        ¦  «        Šˆfd„| j                             ¦   «         D ¦   «         }|sdS  ||‰pd| j        ¬¦  «        }t          j
        d	t          d
|¦  «        › �¦  «         |S )z[Save a cross-dataset bar chart of the per-dataset metric with the mean across all datasets.r   )ÚTASK2METRIC)Úplot_multitrain_resultsr  c                óŠ   •— i | ]?\  }}|¯|t          |                     ‰|                     d d¦  «        ¦  «        ¦  «        “Œ@S )rŽ   ri  )rŸ  r§   )rM   r·   rö  Úkeys      €rP   rn  z-MultiTrainer.plot_results.<locals>.<dictcomp>c  sI   ø€ ÐhÐhÐhÁFÀCÈÐfgÐh�#•u˜QŸUšU 3¨¯ª¨i¸Ñ(=Ô(=Ñ>Ô>Ñ?Ô?ÐhÐhÐhrR   NrŽ   )rè  ro   rá  rc  )Úultralytics.cfgrå  Úultralytics.utils.plottingræ  r§   rb   rl   r   ro   r   r´   r!   )r˜   rå  ræ  ÚscoresÚfnamerè  s        @rP   r-   zMultiTrainer.plot_results]  sÀ   ø€ à/Ð/Ð/Ð/Ð/Ð/ØFÐFÐFÐFÐFÐFà�oŠo˜dœiŸmšm¨FÑ3Ô3Ñ4Ô4ˆØhÐhÐhÐhÈdÌl×N`ÒN`ÑNbÔNbÐhÑhÔhˆØð 	Ø�4Ø'Ð'¨°CÐ4D¸9ÈtÌ}Ð]Ñ]Ô]ˆÝŒÐNµX¸fÀeÑ5LÔ5LÐNÐNÑOÔOÐOØˆrR   r  r²  )r   r³  r´  rµ  r�   r»   rÌ  r-   rK   rR   rP   r·  r·  Ä  sh   € € € € € ðð ð6ð ð ð ð ð$[ð [ð [ðzð ð ðð ð ð ð rR   r·  )_rµ  Ú
__future__r   râ  r  rá   r¶   r‰  r“  r   r   r   r   Ú	functoolsr   Úpathlibr	   r9  r�  rß   r
   rä   r   r   Úultralyticsr   ré  r   r   r   Úultralytics.data.utilsr   r   r   Úultralytics.nn.distill_modelr   Úultralytics.nn.tasksr   Úultralytics.optimr   Úultralytics.utilsr   r   r   r   r   r   r   r   r   r    r!   r"   Úultralytics.utils.autobatchr#   Úultralytics.utils.checksr$   r%   r&   r'   r(   Úultralytics.utils.distr)   r*   Úultralytics.utils.filesr+   rÈ  r,   rê  r-   Úultralytics.utils.torch_utilsr.   r/   r0   r1   r2   r3   r4   r5   r6   r7   r8   r9   r:   r;   r<   r=   r>   r?   r@   rA   rB   rC   rE   r·  rK   rR   rP   ú<module>rû     sè  ððð ð #Ð "Ð "Ð "Ð "Ð "à 	€	€	€	Ø €€€Ø 	€	€	€	Ø Ð Ð Ð Ø €€€Ø €€€Ø Ð Ð Ð Ð Ð Ð Ð Ø (Ð (Ð (Ð (Ð (Ð (Ð (Ð (Ø Ð Ð Ð Ð Ð Ø Ð Ð Ð Ð Ð à Ð Ð Ð Ø €€€Ø %Ð %Ð %Ð %Ð %Ð %Ø Ð Ð Ð Ð Ð Ð Ð à #Ð #Ð #Ð #Ð #Ð #Ø DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DØ iÐ iÐ iÐ iÐ iÐ iÐ iÐ iÐ iÐ iØ :Ð :Ð :Ð :Ð :Ð :Ø 0Ð 0Ð 0Ð 0Ð 0Ð 0Ø #Ð #Ð #Ð #Ð #Ð #ðð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ?Ð >Ð >Ð >Ð >Ð >Ø oÐ oÐ oÐ oÐ oÐ oÐ oÐ oÐ oÐ oÐ oÐ oÐ oÐ oØ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DØ 2Ð 2Ð 2Ð 2Ð 2Ð 2Ø 6Ð 6Ð 6Ð 6Ð 6Ð 6Ø 3Ð 3Ð 3Ð 3Ð 3Ð 3ðð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð4rð rð rð rð rñ rô rð rðj#dð dð dð dð dñ dô dð dð dð drR   