BVh7dZddlZddlmZddlmZddlmZddl m Z ddl m Z ddl m Z dd lmZdd lmZdd lmZdd lmZd ZddZ ddZ ddZ ddZ ddZy)z;Keras training and evaluation routines for eager execution.N) GradientTape)tensor_conversion)backend)training_utils)training_utils_v1)loss_scale_optimizer) losses_utils)math_ops) tf_logging)nestcttj|dz5|||}ddd|S#1swYSxYw)N_loss)r name_scope)outputstargetsloss_fn output_namelosss `/home/dcms/DCMS/lib/python3.12/site-packages/tensorflow/python/keras/engine/training_eager_v1.py_eager_loss_fnr s;+/0% 7G $D% +% +s -7c tj|}tj|}g}|rtt|jt|k7r.|jDcgc]}|dn|j d}}|}|j ||||d|j }|j|jDcgc] }||jvr|j"c}|Scc}wcc}w)aCalculates the metrics for each output of the given model. Args: model: The model on which metrics are being calculated. outputs: The outputs of the given model. targets: The predictions or targets of the given model. sample_weights: Optional list of sample weights for each output. masks: Optional list of masks for each output. Returns: Returns the metric results for each output of the model. NrT)rsample_weightsmasks&return_weighted_and_unweighted_metricsskip_target_masks) r flattenlen_targetspop_handle_metrics_prepare_skip_target_masksextendmetrics_compile_metric_functionsresult) modelrrrrmetric_resultst new_targetsms r_eager_metrics_fnr+&s LL !' LL !'.  5>>c'l*9>45!)$Q /kg**%/3::< +>N}} %11 1hhj 's C*<%C/cd}i}|jr||d<t|dk(rt|ts|d}t dt j |Dr$t jtj|}||fi|}t j |}|rtj||}|rVg} |D]M} | 8| jtjtj| =| jdO| }|D cgc]} t| dd} } t j |}g} tj d5|j"Dcgc]}|| }}|j$}|s|s|r t'dt'd t)|D]\}}|r||nd}| |}tj |j*|d z5|kt-j.|||j0}||}nCt-j.|||j0}t3j4|| \}}}||z}t7|d r|j9||||}t3j:||t2j<j> }|j@}|t2j<jBk(rt2j<jD}t3jF||}n+|||||| }t2j<jD}dddt|jHdkDr| j||t2j<jDk(rt3jJ}||jL|zz }|r+|t3jJt-jN|z }ddd||| | fScc} wcc}w#1swYxYw#1swY%xYw)aCalculates the loss for a given model. Args: model: The model on which metrics are being calculated. inputs: Either a dictionary of inputs to the model or a list of input arrays. targets: List of target arrays. output_loss_metrics: List of metrics that are used to aggregated output loss values. sample_weights: Optional list of sample weight arrays. training: Whether the model should be run in inference or training mode. Returns: Returns the model output, total loss, loss value calculated using the specified loss function and masks for each output. The total loss includes regularization losses and applies masking and sample weighting to the loss value. rtrainingc3fK|])}t|tjttf+yw)N) isinstancenpndarrayfloatint).0input_ts r z_model_loss..vs* + 2::uc23 +s/1N _keras_maskrz?The model cannot be trained because it has no loss to optimize.z@The model cannot be evaluated because it has no loss to compute.r) sample_weight reduction)r9r:)r:)(_expects_training_argrr0dictanyr r map_structurer"convert_to_tensor_v2_with_dispatchr#cast_if_floating_dtype_and_mismatchappendcast_if_floating_dtypegetattrrrloss_functionslosses ValueError enumerate output_namesr castdtyper squeeze_or_expand_dimensionshasattrcallcompute_weighted_loss ReductionV2NONEr:AUTOSUM_OVER_BATCH_SIZEreduce_weighted_lossrscale_loss_for_distribution_loss_weights_listadd_n)r&inputsroutput_loss_metricsrr- total_losskwargsoutsnew_sample_weightsvalr(r output_lossesrloss_fns custom_lossesiweightsmask_per_sample_lossesweighted_lossesloss_reduction output_losss r _model_lossriQs:* &   !F:[Aj6 AYF  +\\&) ++  <? ? )6> 7%3q!g 1Xd   e003g= >%H  tT!W]]3$ _GmmGT!W]];G991 D!W tOG 7K (%ll71:tAw? (>>#$00557/#,,.|77<< <)55IIN%999+  DG7K+'33GG.K%HV U]] a 303K@A <33GG G">>{K E,,Q/+==jm6>rL<< .. '))jQI)T z=% //a : %H%H#I)I)sEO'O=%O,-O,1A/O= EO1(B1O=,O=1O: 6O==Pc 8tj|rdnd5tj|5t 5}t ||||||\}}} } t |jtjr|jj|} n|} ddd|r|j} | rt|dr|j nj | } t |jtjr|jj| } |jj!t#| | nt%j&d  fcdddcdddS#1swYxYw#1swYnxYwdddy#1swYyxYw)aCalculate the loss and gradient for one input batch. The model weights are updated if training is set to True. Args: model: Model whose loss has to be calculated. inputs: List of input arrays. targets: List of target arrays. output_loss_metrics: List of metrics that are used to aggregated output loss values. sample_weights: Optional list of sample weight arrays. training: The boolean represents if the weights of the model are updated. 'fit' methods will set this to True while 'evaluate' methods will set this to False. Returns: output of the model, total loss, the loss and the mask associated with each output. Raises: ValueError: If the model has no loss to optimize. r.r)rXrr-N _backwardszThe list of trainable weights is empty. Make sure that you are not setting model.trainable to False before compiling the model.)reager_learning_phase_scoperRespectCompiledTrainableStaterrir0 optimizerrLossScaleOptimizerget_scaled_losstrainable_weightsrLrkgradientget_unscaled_gradientsapply_gradientsziploggingwarning)r&rWrrXrr-taper[rYr^rscaled_total_lossrqgradss r_process_single_batchr{sv8))x!Q? 22259 2  '4 "5+  !-dJ u EOO%9%L%L M!OO;;JG& '11  5, '   4!2 3-- 13DE% ,??AOO::5AE // ) )#e5F*G H/ 0 ]E 1A 2 2 2 ' ' 2 2 2 2 2s;F E;AE/CE; F/E8 4E;;F FFctj||}t||||d|\}}}}t|ts|g}t |||||} t j|}||| dS)aCalculates the loss and gradient updates for one input batch. Args: model: Model whose loss has to be calculated. inputs: Input batch data. targets: Target batch data. sample_weights: Sample weight batch data. output_loss_metrics: List of metrics that are used to aggregated output loss values. Returns: Dict with three items: 'total_loss': list with a single tensor for overall loss, 'output_losses': list of tensors for loss corresponding to each of the model output. Could be a empty list when model has only one output. 'metrics': list of tensors for metric specified. Trr-rXrrrYr^r#)rcast_to_model_input_dtypesr{r0listr+r r r&rWrrrXr[rYr^rmetrics_resultss rtrain_on_batchr s,  7 7 F&   '1 3)$ M5 D$  6D% T7>H/||J'*"($ &&c *tj||}tjd5t ||||d|\}}}}dddt t s|g}t||||} tj}|| dS#1swYGxYw)arCalculates the loss for one input batch. Args: model: Model whose loss has to be calculated. inputs: Input batch data. targets: Target batch data. sample_weights: Sample weight batch data. output_loss_metrics: List of metrics that are used to aggregated output loss values. Returns: Dict with three items: 'total_loss': single tensor for overall loss, 'output_losses': list of tensors for loss corresponding to each of the model output. Could be a empty list when model has only one output. 'metrics': list of tensors for metric specified. rFr}Nr~r) rrrrlrir0rr+r rrs r test_on_batchrIs,  7 7 F&))!,6   ) 3  5+D*mU6 D$  6D% T7>H/||J'*"($ &&66s B  B)NN)NNF)__doc__numpyr1 tensorflow.python.eager.backproprtensorflow.python.frameworkrtensorflow.python.kerasrtensorflow.python.keras.enginerr'tensorflow.python.keras.mixed_precisionrtensorflow.python.keras.utilsr tensorflow.python.opsr tensorflow.python.platformr rvtensorflow.python.utilr rr+rir{rrrrrswB99+9<H6*<' (\%)# M0f/3)-#( <2D#''+ &&X"&&* )&r