+ j.^RIt^RIHt^RIHt^RIHt^RIHtH t H t H t ]!4'd^RI H t R RRllt!RR ]4t!R R ] 4tRR ltR#)N)center_to_corners_format)is_scipy_available)HungarianMatcher ImageLoss _set_aux_lossgeneralized_box_iou)linear_sum_assignmentc |V^8dQhR\PR\PR\R\R\/#)inputstargets num_boxesalphagamma)torchTensorintfloat)formats"f/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/loss/loss_grounding_dino.py __annotate__rsB$"$" LL$" \\$"$"  $"  $"cnVP4p\PPWRR7pWQ,^V, ^V, ,,pV^V, V,,pV^8d-W1,^V, ^V, ,,p W,pVP 4V, #)a Loss used in RetinaNet for dense detection: https://huggingface.co/papers/1708.02002. Args: inputs (`torch.FloatTensor` of arbitrary shape): The predictions for each example. targets (`torch.FloatTensor` with the same shape as `inputs`) A tensor storing the binary classification label for each element in the `inputs` (0 for the negative class and 1 for the positive class). num_boxes (`int`): The total number of boxes in the batch. alpha (`float`, *optional*, defaults to 0.25): Optional weighting factor in the range (0,1) to balance positive vs. negative examples. gamma (`int`, *optional*, defaults to 2): Exponent of the modulating factor (1 - p_t) to balance easy vs hard examples. Returns: Loss tensor none) reduction)sigmoidnn functional binary_cross_entropy_with_logitssum) r r rrrprobce_lossp_tlossalpha_ts &&&&& rsigmoid_focal_lossr&s4 >> Dmm<)p)dtype)shapeflattenrrcatzipr logtcdistrr bbox_cost class_cost giou_costviewcpulen enumeratesplitr as_tensorint64)selfoutputsr batch_size num_queriesout_probout_bboxr, label_maptargetv target_bboxrrneg_cost_classpos_cost_classr>r=r? cost_matrixsizesicindicesjs&&& rforward%GroundingDinoHungarianMatcher.forwardDsg,#*("3"9"9""= 8$,,Q2::<<(00A6\* YY[^_i[st[sFWi *@ A A[stu ..R."FF iiW =W7W => e)8a(lT>Q=V=V=XVW$5G KK; ))A()KMefqMrss nny04??Z3OORVR`R`clRll !&&zCGGI *12'QQwZ'2;D[EVEVW\^`Ea;bc;b41a(1.;bckrskrcgcdfg%++6QVQ\Q\8]^krss7 u !>(3cssK KK K%9AK+N) __name__ __module__ __qualname____firstlineno__rno_gradrY__static_attributes____classdictcell__ __classdict__s@rr(r(Cs ]]_8t8trr(cda]tRt^toRtRt]P!4R4tRt Rt Rt Vt R#)GroundingDinoImageLossaZ This class computes the losses for `GroundingDinoForObjectDetection`. The process happens in two steps: 1) we compute hungarian assignment between ground truth boxes and the outputs of the model 2) we supervise each pair of matched ground-truth / prediction (supervise class and box). Args: matcher (`GroundingDinoHungarianMatcher`): Module able to compute a matching between targets and proposals. focal_alpha (`float`): Alpha parameter in focal loss. losses (`list[str]`): List of all the losses to be applied. See `get_loss` for a list of all available losses. ch\PPV4WnW nW0nR#N)rModule__init__matcher focal_alphalosses)rGrkrlrms&&&&rrjGroundingDinoImageLoss.__init__s$ 4  & rc  VR,pVPp\P!VUu.uFp\VR,4NK upVR7pVP 4P R4P R8P^4p \PPV P4VP44p RV /p V #uupi)z Compute the cardinality error, i.e. the absolute error in the number of predicted non-empty boxes. This is not really a loss, it is intended for logging purposes only. It doesn't propagate gradients. r*r-)deviceg?cardinality_errorr5) rprrErBrmaxvaluesr rrl1_lossr) rGrHr rWrr*rprOtarget_lengths card_predcard_errrms &&&&& rloss_cardinality'GroundingDinoImageLoss.loss_cardinalitys"')R'Q#a.?*@')R[ab^^%))"-44s:??B ==(():NR&W^R&WnR&RpVP'd#\Wx4pVF pVVR&VVR&K WR&V !W4oVP'dQRV RV RVRV/pV !VV4pVP4UUu/uFwppVR,VbK pppSPV4R R R VPR VP/oVP'd>SP4UUu/uFwppVR,VbK pppSPV4VP'dv/p\VP^, 4FBpTPSP4UUu/uFwppVR V 2,VbK upp4KD SPV4\!VV3RlS44pVSV3#uuppiuuppiuuppi))r>r=r?)rkrlrmr*r+r,rNauxiliary_outputs_encrr2 loss_bbox loss_giourc3b<"TF$qS9gK SV,SV,,xK& R#5irhr[).0k loss_dict weight_dicts& r 6GroundingDinoForObjectDetectionLoss..s)Ti CS,y|k!n,,is//)labelsr0 cardinality)r(r>r=r?rfrlr~auxiliary_lossr two_stageitemsupdatebbox_loss_coefficientgiou_loss_coefficientrangedecoder_layersr )r*rrpr+configr,r outputs_class outputs_coordencoder_logitsencoder_pred_boxesrkrm criterion outputs_lossr aux_outputencoder_outputs_lossencoder_loss_dictrrOenc_weight_dictaux_weight_dictrUr$rrs&&&&&&&&&&& @@r#GroundingDinoForObjectDetectionLossrsM,$$0@0@FL\L\G0F&&&I LLL#!+!+ ) )-G+J'1J| $&/J{ #,->(),/I  n , *    &&:FC7H7N7N7PQ7Ptq!QZ]7PQ*+ 3V11V11K 5@5F5F5HI5HTQ1v:q=5HI?+ v,,q01A  " "{?P?P?R#S?Rtq!A!A3KN?R#S T2?+ TiT TD - --)RJ $Ts$HH! H' )r1r )NNNN)rtorch.nnrimage_transformsrutilsrloss_for_object_detectionrrrrscipy.optimizer r&r(rfrr[rrrsR 7&ff4 $"N:t$4:tzVYVrF.r