§
    “©ºj‹+  ã                   ó  — d dl Z d dlmZmZmZmZmZ d dlZd dlm	Z	 d dl
mZ d dlmZ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mZmZ d d	lmZmZ d d
l m!Z!m"Z" erd dl#m$Z$m%Z%  G d„ de¦  «        Z&deeef         dej'        fd„ZdS )é    N)ÚTYPE_CHECKINGÚAnyÚCallableÚOptionalÚUnion)Úoverride)Ú	Precision)Ú
ThroughputÚget_available_flops)Ú_plugin_to_compute_dtype)ÚCallback)	ÚBitsandbytesPrecisionÚDeepSpeedPrecisionÚDoublePrecisionÚFSDPPrecisionÚHalfPrecisionÚMixedPrecisionr	   ÚTransformerEnginePrecisionÚXLAPrecision)ÚRunningStageÚ	TrainerFn)Úrank_zero_onlyÚrank_zero_warn)ÚLightningModuleÚTrainerc                   ó  ‡ — e Zd ZdZ	 ddeegef         deeegef                  deddfˆ fd„Ze	dd	d
dde
ddfd„¦   «         Zd d„Z ej        ¦   «         dd	d
ddededdf
d„¦   «         Zddd	dee         ddfd„Ze	edd	deddfd„¦   «         ¦   «         Ze	edd	d
ddedededdfd„¦   «         ¦   «         Ze	edd	deddfd„¦   «         ¦   «         Ze	edd	d
ddededededdfd„¦   «         ¦   «         Ze	edd	deddfd„¦   «         ¦   «         Ze	edd	deddfd„¦   «         ¦   «         Ze	edd	d
ddededededdfd„¦   «         ¦   «         Ze	edd	deddfd„¦   «         ¦   «         Ze	edd	d
ddededededdfd„¦   «         ¦   «         Zˆ xZS )!ÚThroughputMonitoraÌ  Computes and logs throughput with the :class:`~lightning_fabric.utilities.throughput.Throughput`

    Example::

        class MyModel(LightningModule):
            def setup(self, stage):
                with torch.device("meta"):
                    model = MyModel()

                    def sample_forward():
                        batch = torch.randn(..., device="meta")
                        return model(batch)

                    self.flops_per_batch = measure_flops(model, sample_forward, loss_fn=torch.Tensor.sum)


        logger = ...
        throughput = ThroughputMonitor(batch_size_fn=lambda batch: batch.size(0))
        trainer = Trainer(max_steps=1000, log_every_n_steps=10, callbacks=throughput, logger=logger)
        model = MyModel()
        trainer.fit(model)

    Notes:
        - It assumes that the batch size is the same during all iterations.
        - It will try to access a ``flops_per_batch`` attribute on your ``LightningModule`` on every iteration.
          We suggest using the :func:`~lightning_fabric.utilities.throughput.measure_flops` function for this.
          You might want to compute it differently each time based on your setup.

    Args:
        batch_size_fn: A function to compute the number of samples given a batch.
        length_fn: A function to compute the number of items in a sample given a batch.
        \**kwargs: See available parameters in
            :class:`~lightning_fabric.utilities.throughput.Throughput`

    NÚbatch_size_fnÚ	length_fnÚkwargsÚreturnc                 óÔ   •— t          ¦   «                              ¦   «          || _        || _        || _        d | _        i | _        i | _        i | _        i | _	        i | _
        d | _        d S ©N)ÚsuperÚ__init__r    r   r   Úavailable_flopsÚ_throughputsÚ_t0sÚ_lengthsÚ_samplesÚ_batchesÚ_module_has_flops)Úselfr   r   r    Ú	__class__s       €ú~/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/pytorch_lightning/callbacks/throughput_monitor.pyr%   zThroughputMonitor.__init__O   sk   ø€ õ 	‰Œ×ÒÑÔÐØˆŒØ*ˆÔØ"ˆŒØ.2ˆÔØ<>ˆÔØ/1ˆŒ	Ø13ˆŒØ13ˆŒØ13ˆŒØ.2ˆÔÐÐó    Útrainerr   Ú	pl_moduler   Ústagec                 ón  — t          |j        ¦  «        }t          |j        j        |¦  «        | _        |t          j        k    r9|j        r2t          d| j        |j
        dœ| j        ¤Ž}|| j        t          j        <   t          d| j        |j
        dœ| j        ¤Ž}|j        j        }|€J ‚|| j        |<   d S )N)r&   Ú
world_size© )r   Úprecision_pluginr   ÚstrategyÚroot_devicer&   r   ÚFITTINGÚenable_validationr
   r5   r    r'   r   Ú
VALIDATINGÚstater3   )r-   r1   r2   r3   ÚdtypeÚ
throughputs         r/   ÚsetupzThroughputMonitor.setup^   sÄ   € å(¨Ô)AÑBÔBˆÝ2°7Ô3CÔ3OÐQVÑWÔWˆÔà•IÔ%Ò%Ð%¨'Ô*CÐ%å#Ðw°DÔ4HÐU\ÔUgÐwÐwÐkoÔkvÐwÐwˆJØ9CˆDÔ�lÔ5Ñ6åÐs°Ô0DÐQXÔQcÐsÐsÐgkÔgrÐsÐsˆ
Ø”Ô#ˆØÐ Ð Ð Ø#-ˆÔ˜%Ñ Ð Ð r0   c                 ó   — |j         j        }|€J ‚|j         j        t          j        k    p|| j        v}|r=| j        |                              ¦   «          d| j        |<   d| j        |<   d| j	        |<   t          j        ¦   «         | j        |<   d S )Nr   )r=   r3   Úfnr   r:   r*   r'   Úresetr)   r+   ÚtimeÚperf_counterr(   )r-   r1   r3   Úreset_neededs       r/   Ú_startzThroughputMonitor._startm   s—   € Ø”Ô#ˆØÐ Ð Ð à”}Ô'­9Ô+<Ò<ÐZÀÈTÌ]Ð@Zˆàð 	%ØÔ˜eÔ$×*Ò*Ñ,Ô,Ð,Ø#$ˆDŒM˜%Ñ Ø#$ˆDŒM˜%Ñ Ø#$ˆDŒM˜%Ñ åÔ,Ñ.Ô.ˆŒ	�%ÑÐÐr0   ÚbatchÚiter_numc                 óø  — |j         j        }|€J ‚| j        |         }|j        j        j        dk    rt          j                             ¦   «          t          j
        ¦   «         | j        |         z
  }| j        �(| j        |xx         |                      |¦  «        z  cc<   | j        €At          |d¦  «        | _        | j        s%t!          dt          |¦  «        j        › d�¦  «         | j        r|j        nd }| j        |xx         |                      |¦  «        z  cc<   | j        |xx         dz  cc<   |                     || j        |         | j        |         | j        €d n| j        |         |¬¦  «         d S )NÚcudaÚflops_per_batchzdWhen using the `ThroughputMonitor`, you need to define a `flops_per_batch` attribute or property in z to compute the FLOPs.é   )rD   ÚbatchesÚsamplesÚlengthsÚflops)r=   r3   r'   r8   r9   ÚtypeÚtorchrK   ÚsynchronizerD   rE   r(   r   r)   r,   Úhasattrr   Ú__name__rL   r*   r   r+   Úupdate)	r-   r1   r2   rH   rI   r3   r?   ÚelapsedrL   s	            r/   Ú_updatezThroughputMonitor._update{   s®  € à”Ô#ˆØÐ Ð Ð ØÔ& uÔ-ˆ
àÔÔ'Ô,°Ò6Ð6åŒJ×"Ò"Ñ$Ô$Ð$åÔ#Ñ%Ô%¨¬	°%Ô(8Ñ8ˆØŒ>Ð%ØŒM˜%Ð Ð Ô  D§N¢N°5Ñ$9Ô$9Ñ9Ð Ð Ñ àÔ!Ð)Ý%,¨YÐ8IÑ%JÔ%JˆDÔ"ØÔ)ð ÝðLÝ 	™?œ?Ô3ðLð Lð Lñô ð ð
 8<Ô7MÐW˜)Ô3Ð3ÐSWˆàŒ�eÐÐÔ × 2Ò 2°5Ñ 9Ô 9Ñ9ÐÐÑØŒ�eÐÐÔ Ñ!ÐÐÑà×ÒØØ”M %Ô(à”M %Ô(Ø œNÐ2�D�D¸¼ÀeÔ8LØ!ð 	ñ 	
ô 	
ð 	
ð 	
ð 	
r0   c                 ó  ‡‡— |j         j        sd S |j        j        Š‰€J ‚| j        ‰         Š‰                     ¦   «         }ˆˆfd„|                     ¦   «         D ¦   «         }|j                              ||¬¦  «         d S )Nc                 ó<   •— i | ]\  }}‰j         › ‰j        › |› �|“ŒS r6   )ÚvalueÚ	separator)Ú.0ÚkÚvr3   r?   s      €€r/   ú
<dictcomp>z.ThroughputMonitor._compute.<locals>.<dictcomp>§   s4   ø€ Ð\Ð\Ð\ÁDÀAÀq�e”kÐ< :Ô#7Ð<¸Ð<Ð<¸aÐ\Ð\Ð\r0   )Ústep)Ú_logger_connectorÚshould_update_logsr=   r3   r'   ÚcomputeÚitemsÚlog_metrics)r-   r1   rI   Úmetricsr3   r?   s       @@r/   Ú_computezThroughputMonitor._computeŸ   s‘   øø€ ØÔ(Ô;ð 	ØˆFØ”Ô#ˆØÐ Ð Ð ØÔ& uÔ-ˆ
Ø×$Ò$Ñ&Ô&ˆà\Ð\Ð\Ð\Ð\ÈGÏMÊMÉOÌOÐ\Ñ\Ô\ˆØÔ!×-Ò-¨g¸HÐ-ÑEÔEÐEÐEÐEr0   Ú_c                 ó0   — |                       |¦  «         d S r#   ©rG   ©r-   r1   rj   s      r/   Úon_train_startz ThroughputMonitor.on_train_startª   ó   € ð 	�Š�GÑÔÐÐÐr0   Úoutputsc                 ó°   — |                       ||||j        j        dz   ¦  «         |j                             ¦   «         s|                      |¦  «         d S d S )NrM   )rY   Úfit_loopÚtotal_batch_idxÚ_should_accumulateri   )r-   r1   r2   rp   rH   rj   s         r/   Úon_train_batch_endz$ThroughputMonitor.on_train_batch_end¯   sc   € ð
 	�Š�W˜i¨°Ô0@Ô0PÐSTÑ0TÑUÔUÐUð Ô×2Ò2Ñ4Ô4ð 	#Ø�MŠM˜'Ñ"Ô"Ð"Ð"Ð"ð	#ð 	#r0   c                 óB   — |j         rd S |                      |¦  «         d S r#   )Úsanity_checkingrG   rm   s      r/   Úon_validation_startz%ThroughputMonitor.on_validation_startº   s,   € ð Ô"ð 	ØˆFØ�Š�GÑÔÐÐÐr0   Ú__c                 ó    — |j         rd S |j        j        j        j        }|                      ||||¦  «         |                      ||¦  «         d S r#   )rw   Ú_evaluation_loopÚbatch_progressÚtotalÚreadyrY   ri   ©r-   r1   r2   rp   rH   rj   ry   rI   s           r/   Úon_validation_batch_endz)ThroughputMonitor.on_validation_batch_endÁ   sW   € ð
 Ô"ð 	ØˆFØÔ+Ô:Ô@ÔFˆØ�Š�W˜i¨°Ñ9Ô9Ð9Ø�Š�g˜xÑ(Ô(Ð(Ð(Ð(r0   c                 óœ  — |j         s|j        j        t          j        k    rd S | j        t          j                 j        }| j        t          j	                 j        }|r|d         nd}|r|d         nd}| j
        t          j                 |z   }| j
        t          j	                 |z
  }|}	| j
        t          j        xx         ||	z   z  cc<   d S )Néÿÿÿÿg        )rw   r=   rB   r   r:   r'   r   ÚTRAININGÚ_timer<   r(   )
r-   r1   rj   Útrain_timesÚ	val_timesÚtrain_elapsedÚval_elapsedÚtraining_finishedÚtime_between_train_and_valÚval_times
             r/   Úon_validation_endz#ThroughputMonitor.on_validation_endÌ   sÏ   € ð Ô"ð 	 g¤mÔ&6½)Ô:KÒ&KÐ&KØˆFàÔ'­Ô(=Ô>ÔDˆØÔ%¥lÔ&=Ô>ÔDˆ	à+6Ð?˜ Bœ˜¸CˆØ'0Ð9�i ”m�m°cˆð !œI¥lÔ&;Ô<¸}ÑLÐØ%)¤Y­|Ô/FÔ%GÐJ[Ñ%[Ð"ØˆØŒ	•,Ô'Ð(Ð(Ô(Ð,FÈÑ,QÑQÐ(Ð(Ñ(Ð(Ð(r0   c                 ó0   — |                       |¦  «         d S r#   rl   rm   s      r/   Úon_test_startzThroughputMonitor.on_test_startÞ   ro   r0   c                 óŽ   — |j         j        j        j        }|                      ||||¦  «         |                      ||¦  «         d S r#   )r{   r|   r}   r~   rY   ri   r   s           r/   Úon_test_batch_endz#ThroughputMonitor.on_test_batch_endã   sG   € ð
 Ô+Ô:Ô@ÔFˆØ�Š�W˜i¨°Ñ9Ô9Ð9Ø�Š�g˜xÑ(Ô(Ð(Ð(Ð(r0   c                 ó0   — |                       |¦  «         d S r#   rl   rm   s      r/   Úon_predict_startz"ThroughputMonitor.on_predict_startì   ro   r0   c                 óŽ   — |j         j        j        j        }|                      ||||¦  «         |                      ||¦  «         d S r#   )Úpredict_loopr|   r}   r~   rY   ri   r   s           r/   Úon_predict_batch_endz&ThroughputMonitor.on_predict_batch_endñ   sG   € ð
 Ô'Ô6Ô<ÔBˆØ�Š�W˜i¨°Ñ9Ô9Ð9Ø�Š�g˜xÑ(Ô(Ð(Ð(Ð(r0   r#   )r1   r   r!   N)rV   Ú
__module__Ú__qualname__Ú__doc__r   r   Úintr   r%   r   Ústrr@   rG   rS   Úinference_moderY   ri   r   rn   ru   rx   r€   rŒ   rŽ   r�   r’   r•   Ú__classcell__)r.   s   @r/   r   r   *   s4  ø€ € € € € ð"ð "ðJ `dð3ð 3Ø% s e¨S jÔ1ð3Ø>FÀxÐQTÐPUÐWZÐPZÔG[Ô>\ð3Øorð3à	ð3ð 3ð 3ð 3ð 3ð 3ð ð.˜Yð .Ð3Dð .ÈSð .ÐUYð .ð .ð .ñ „Xð.ð/ð /ð /ð /ð €UÔÑÔð!
˜yð !
Ð5Fð !
Èsð !
Ð^að !
Ðfjð !
ð !
ð !
ñ Ôð!
ðF	Fð 	F 	ð 	F°X¸c´]ð 	FÈdð 	Fð 	Fð 	Fð 	Fð Øð ið °Sð ¸Tð ð ð ñ „^ñ „Xðð Øð#Ø ð#Ø->ð#ØILð#ØUXð#Ø^að#à	ð#ð #ð #ñ „^ñ „Xð#ð Øð¨9ð ¸#ð À$ð ð ð ñ „^ñ „Xðð
 Øð)Ø ð)Ø->ð)ØILð)ØUXð)Ø^að)Øilð)à	ð)ð )ð )ñ „^ñ „Xð)ð ØðR¨ð R¸ð RÀð Rð Rð Rñ „^ñ „XðRð  Øð Yð °Cð ¸Dð ð ð ñ „^ñ „Xðð Øð)Ø ð)Ø->ð)ØILð)ØUXð)Ø^að)Øilð)à	ð)ð )ð )ñ „^ñ „Xð)ð Øð¨	ð °sð ¸tð ð ð ñ „^ñ „Xðð Øð)Ø ð)Ø->ð)ØILð)ØUXð)Ø^að)Øilð)à	ð)ð )ð )ñ „^ñ „Xð)ð )ð )ð )ð )r0   r   Úpluginr!   c                 ó®  — t          | t          ¦  «        st          | ¦  «        S t          | t          ¦  «        r| j        S t          | t
          ¦  «        r| j        S t          | t          ¦  «        r#| j        dk    rt          j
        nt          j        S t          | t          ¦  «        rt          j        S t          | t          t          f¦  «        r| j        S t          | t"          ¦  «        rt          j        S t          | t&          ¦  «        r| j        j        pt          j        S t          | t          ¦  «        rt          j        S t/          | ¦  «        ‚)Nz
bf16-mixed)Ú
isinstancer	   Úfabric_plugin_to_compute_dtyper   r>   r   Ú_desired_input_dtyper   Ú	precisionrS   Úbfloat16Úhalfr   Údoubler   r   Ú_desired_dtyper   Úint8r   Úmixed_precision_configÚreduce_dtypeÚfloat32ÚNotImplementedError)r�   s    r/   r   r   û   s*  € å�f�iÑ(Ô(ð 6Ý-¨fÑ5Ô5Ð5Ý�&Õ/Ñ0Ô0ð ØŒ|ÐÝ�&�-Ñ(Ô(ð +ØÔ*Ð*Ý�&�.Ñ)Ô)ð RØ!'Ô!1°\Ò!AÐ!A�uŒ~ˆ~ÅuÄzÐQÝ�&�/Ñ*Ô*ð ÝŒ|ÐÝ�&�<Õ);Ð<Ñ=Ô=ð %ØÔ$Ð$Ý�&Õ4Ñ5Ô5ð ÝŒzÐÝ�&�-Ñ(Ô(ð KØÔ,Ô9ÐJ½U¼]ÐJÝ�&�)Ñ$Ô$ð ÝŒ}ÐÝ
˜fÑ
%Ô
%Ð%r0   )(rD   Útypingr   r   r   r   r   rS   Útyping_extensionsr   Úlightning_fabric.pluginsr	   ÚFabricPrecisionÚ%lightning_fabric.utilities.throughputr
   r   r   r    Úpytorch_lightning.callbacksr   Úpytorch_lightning.pluginsr   r   r   r   r   r   r   r   Ú pytorch_lightning.trainer.statesr   r   Ú%pytorch_lightning.utilities.rank_zeror   r   Úpytorch_lightningr   r   r   r>   r6   r0   r/   ú<module>r¶      sÏ  ðð €€€Ø @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @à €€€Ø &Ð &Ð &Ð &Ð &Ð &à AÐ AÐ AÐ AÐ AÐ AØ QÐ QÐ QÐ QÐ QÐ QÐ QÐ QØ lÐ lÐ lÐ lÐ lÐ lØ 0Ð 0Ð 0Ð 0Ð 0Ð 0ð
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð EÐ DÐ DÐ DÐ DÐ DÐ DÐ DØ PÐ PÐ PÐ PÐ PÐ PÐ PÐ Pàð ;Ø:Ð:Ð:Ð:Ð:Ð:Ð:Ð:ðN)ð N)ð N)ð N)ð N)˜ñ N)ô N)ð N)ðb& U¨?¸IÐ+EÔ%Fð &È5Ì;ð &ð &ð &ð &ð &ð &r0   