0hBdZddlmcmZddlmZddlmZddl m Z ddl m Z e dgGd d Z dd Zdd Zd ZdZe j$fdZe dgde j$dfdZdZdZdZdZdZy)z$Utilities related to loss functions.N)backend) keras_tensor)tf_utils) keras_exportzkeras.losses.Reduction)v1c@eZdZdZdZdZdZdZedZ edZ y) ReductionV2aTypes of loss reduction. Contains the following values: * `AUTO`: Indicates that the reduction option will be determined by the usage context. For almost all cases this uses `SUM_OVER_BATCH_SIZE`. When used with `tf.distribute.Strategy`, outside of built-in training loops such as `tf.keras` `compile` and `fit`, we expect reduction value to be `SUM` or `NONE`. Using `AUTO` in that case will raise an error. * `NONE`: No **additional** reduction is applied to the output of the wrapped loss function. When non-scalar losses are returned to Keras functions like `fit`/`evaluate`, the unreduced vector loss is passed to the optimizer but the reported loss will be a scalar value. Caution: **Verify the shape of the outputs when using** `Reduction.NONE`. The builtin loss functions wrapped by the loss classes reduce one dimension (`axis=-1`, or `axis` if specified by loss function). `Reduction.NONE` just means that no **additional** reduction is applied by the class wrapper. For categorical losses with an example input shape of `[batch, W, H, n_classes]` the `n_classes` dimension is reduced. For pointwise losses you must include a dummy axis so that `[batch, W, H, 1]` is reduced to `[batch, W, H]`. Without the dummy axis `[batch, W, H]` will be incorrectly reduced to `[batch, W]`. * `SUM`: Scalar sum of weighted losses. * `SUM_OVER_BATCH_SIZE`: Scalar `SUM` divided by number of elements in losses. This reduction type is not supported when used with `tf.distribute.Strategy` outside of built-in training loops like `tf.keras` `compile`/`fit`. You can implement 'SUM_OVER_BATCH_SIZE' using global batch size like: ``` with strategy.scope(): loss_obj = tf.keras.losses.CategoricalCrossentropy( reduction=tf.keras.losses.Reduction.NONE) .... loss = tf.reduce_sum(loss_obj(labels, predictions)) * (1. / global_batch_size) ``` Please see the [custom training guide]( https://www.tensorflow.org/tutorials/distribute/custom_training) for more details on this. autononesumsum_over_batch_sizec^|j|j|j|jfSN)AUTONONESUMSUM_OVER_BATCH_SIZE)clss Y/var/www/html/engine/venv/lib/python3.12/site-packages/tf_keras/src/utils/losses_utils.pyallzReductionV2.allQs##((CGGS-D-DEEch||jvr td|d|jdy)NzInvalid Reduction Key: z. Expected keys are "")r ValueError)rkeys rvalidatezReductionV2.validateUs; cggi )#.CCGGI;aP  rN) __name__ __module__ __qualname____doc__rrrr classmethodrrrrr r sG,\ D D C/FFrr c<tj|xsd5tjst j tjst j j }|j}j }|j}||||z }||dzk(r6|jdjdrt jdgn=||dz k(r5|jdjdrt jdgfcdddSt jt jz }||jdjdr4t jt j|dz|fdfd||jdjdr4t jt j|dz |fdfdfcdddS#1swYyxYw) aHSqueeze last dim if ranks differ from expected by exactly 1. In the common case where we expect shapes to match, `expected_rank_diff` defaults to 0, and we squeeze the last dimension of the larger rank if they differ by 1. But, for example, if `labels` contains class IDs and `predictions` contains 1 probability per class, we expect `predictions` to have 1 more dimension than `labels`, so `expected_rank_diff` would be 1. In this case, we'd squeeze `labels` if `rank(predictions) - rank(labels) == 0`, and `predictions` if `rank(predictions) - rank(labels) == 2`. This will use static shape if available. Otherwise, it will add graph operations, which could result in a performance hit. Args: labels: Label values, a `Tensor` whose dimensions match `predictions`. predictions: Predicted values, a `Tensor` of arbitrary dimensions. expected_rank_diff: Expected result of `rank(predictions) - rank(labels)`. name: Name of the op. Returns: Tuple of `labels` and `predictions`, possibly with last dim squeezed. remove_squeezable_dimensionsNc2tjdgSNr&tfsqueeze predictionssrz.remove_squeezable_dimensions..s ;5rcSrr"r,srr.z.remove_squeezable_dimensions..s rc2tjdgSr(r)labelssrr.z.remove_squeezable_dimensions..s 6B40rcSrr"r1srr.z.remove_squeezable_dimensions..sr)r name_scoperis_tensor_or_extension_typer*convert_to_tensorshapendimsdimsis_compatible_withr+rankcondequal) r2r-expected_rank_diffnamepredictions_shapepredictions_rank labels_shape labels_rank rank_diffs `` rr$r$]s6   DB$B C(#33K@..{;K33F;))&1F'--,22|| "((  #*:*F(;6I.227H7M7M8  #8$!jjrd; 01449J9J:  #:$FRD1;&)(#(#.GGK(2776?:  $  " "2 & 9 9! <''+a/;5#K     b ! 4 4Q 7WW+a/;0F {"Q(#(#(#sD H2CHHc j}|j}ωj}|j}|"| ||z dk7s|ddk(rt\ntjtjz fdtj dtjd fd}tj tj d|\fSj}|j} | dk(rfS|F| D| |z dk(rtjdgn|| z dk(rtjdgfStj} | tjz fdfd fd} tj tj | dfd | fS) aSqueeze or expand last dimension if needed. 1. Squeezes last dim of `y_pred` or `y_true` if their rank differs by 1 (using `remove_squeezable_dimensions`). 2. Squeezes or expands last dim of `sample_weight` if its rank differs by 1 from the new rank of `y_pred`. If `sample_weight` is scalar, it is kept scalar. This will use static shape if available. Otherwise, it will add graph operations, which could result in a performance hit. Args: y_pred: Predicted values, a `Tensor` of arbitrary dimensions. y_true: Optional label `Tensor` whose dimensions match `y_pred`. sample_weight: Optional weight scalar or `Tensor` whose dimensions match `y_pred`. Returns: Tuple of `y_pred`, `y_true` and `sample_weight`. Each of them possibly has the last dimension squeezed, `sample_weight` could be extended by one dimension. If `sample_weight` is None, (y_pred, y_true) is returned. r%r&ctSr)r$y_predy_truesrr.z.squeeze_or_expand_dimensions..s#?#Orc:tjfdS)Nc fSrr"rGsrr.z@squeeze_or_expand_dimensions....s ff5Er)r*r<) is_last_dim_1 squeeze_dimsrHrIsrr.z.squeeze_or_expand_dimensions..s|-E*rrc2tjdgSr(r) sample_weightsrr.z.squeeze_or_expand_dimensions..sBJJ}rd$Crcjfd}tjtjd|fdS)Nc2tjdgSr()r* expand_dimsrOsrr.zMsqueeze_or_expand_dimensions.._maybe_expand_weights..s t!Drr&cSrr"rOsrr.zMsqueeze_or_expand_dimensions.._maybe_expand_weights..s]rr*r<r=)expand_weightsrDrPs r_maybe_expand_weightsz;squeeze_or_expand_dimensions.._maybe_expand_weightss,Dww HHY #^5J  rcZtjtjdS)Nr%rU)rWmaybe_squeeze_weightsrDsr_maybe_adjust_weightsz;squeeze_or_expand_dimensions.._maybe_adjust_weightss(ww HHY "$9;P  rcSrr"rOsrr.z.squeeze_or_expand_dimensions..s r) r7r8r$r*r;r=r<r+rS)rHrIrP y_pred_shape y_pred_rank y_true_shape y_true_rankmaybe_squeeze_dims weights_shape weights_rankweights_rank_tensorrZrWrLrYrDrMs``` @@@@@rsqueeze_or_expand_dimensionsrds0< + % *JJ}rd;M < '1 ,NN=2$?Mv},,''-0#bggfo5IC  GG $a(M 6= ((rcptj|}tjj||dS)a:Computes a safe mean of the losses. Args: losses: `Tensor` whose elements contain individual loss measurements. num_present: The number of measurable elements in `losses`. Returns: A scalar representing the mean of `losses`. If `num_present` is zero, then zero is returned. valuer?)r* reduce_summath divide_no_nan)losses num_present total_losss r _safe_meanrns.v&J 77 [w GGrctjd5}tjtj|||j cdddS#1swYyxYw)z3Computes the number of elements in `losses` tensor. num_elementsrg)dtypeN)rr4r*castsizerq)rkscopes r _num_elementsru sH   N +HuwwrwwvE2&,,GHHHs 6AAc|tjk(r|}|Stj|}|tjk(rt |t |}|S)z2Reduces the individual weighted loss measurements.)r rr*rhrrnru)weighted_losses reductionlosss rreduce_weighted_lossrzsRK$$$ K}}_-  77 7dM/$BCD Krz/keras.__internal__.losses.compute_weighted_lossctj||tjk(rtj}|d}t j |xsd5|t jjj_ t|tjt jfst j|}t|tjt jfst j|}|j j"s%|j }t j$|d}d}nd}t j$||j }t'|d|\}}}t j(||}t+||}|rt j$|}|cdddS#1swYyxYw)aComputes the weighted loss. Args: losses: `Tensor` of shape `[batch_size, d1, ... dN]`. sample_weight: Optional `Tensor` whose rank is either 0, or the same rank as `losses`, or be broadcastable to `losses`. reduction: (Optional) Type of `tf.keras.losses.Reduction` to apply to loss. Default value is `SUM_OVER_BATCH_SIZE`. name: Optional name for the op. Raises: ValueError: If the shape of `sample_weight` is not compatible with `losses`. Returns: Weighted loss `Tensor` of the same type as `losses`. If `reduction` is `NONE`, this has the same shape as `losses`; otherwise, it is scalar. N? weighted_lossfloat32TF)r rrrrr4r*compatrget_default_graph_last_loss_reduction isinstancer KerasTensor RaggedTensorr6rq is_floatingrrrdmultiplyrz) rkrPrxr? input_dtype input_casted_rwrys rcompute_weighted_lossr sk2#K$$$33    D3O 4$AJ &&(=&<#;#;R__"MN))&1F L44booF 00?M||'' ,,KWWVY/FL L v||<  )} E   ++fm<$OY? 774-DI$$$s EF55F>cptjjj}|dkDr|d|z z}|S)zBScales and returns the given loss value by the number of replicas.r%r|)r* distribute get_strategynum_replicas_in_sync) loss_value num_replicass rscale_loss_for_distributionrhs7==--/DDLacL(( rcZd}|D]y}|jjrG|#|jj|jkDr |j}n|j|hddhk(rd}|jjsw|cS|r#|Dcgc]}t j ||}}|Scc}w)arCast a list of losses to a common dtype. If any loss is floating-point, they will all be casted to the most-precise floating-point loss. Otherwise the losses are not casted. We also skip casting losses if there are any complex losses. Args: losses: A list of losses. Returns: `losses`, but they have been casted to a common dtype. Nbfloat16float16r~)rqrrs is_complexr*rr)rk highest_floatrys rcast_losses_to_common_dtyperpsM  :: ! !$ -:L:L(L $ **m,Y0GG ) ::   ;AB4"''$ .BB MCsB(ct|ddS)z"Returns TF-Keras mask from tensor. _keras_maskN)getattr)y_ps rget_maskrs 3 t ,,rc|\tj||j}|8tj||j}t||\}}}||z}|S|}|S)z2Applies any mask on predictions to sample weights.rO)r*rrrqrd)rswmaskrs r apply_maskrsc wwtSYY' >TZZ(B6t2NKD!R $JB IB IrcD|tj||j}|tjtj fvrPtjtj ||j}tj|}|||z z}t|||S)z;Redistribute sample weights considering only valid entries.) r*rrrqr rrrsrhr)rkrrrxtotalvalids rapply_valid_maskrs~ wwtV\\* ));+J+JK KGGBGGDM6<<8EMM$'E EEM !D fb$ ''r)rN)NN)r tensorflow.compat.v2rv2r* tf_keras.srcrtf_keras.src.enginertf_keras.src.utilsr tensorflow.python.util.tf_exportrr r$rdrnrurrzrrrrrrr"rrrs +!! ,':&2.==/=B59C#LX)v HH +>> ?BG-- DHDN:- (r