+ juRt^RIHt^RIt^RIHt^RIHtHtHt^RI H t ^RI H t ^RIHt^RIHtHtHtHtHtHt^R IHtHt^R IHt^R IHt^R IHtH t H!t!H"t"^R I#H$t$^RI%H&t&^RI'H(t(]"PR!]*4t+!RR]PX4t-!RR]PX4t.!RR]PX4t/!RR]PX4t0!RR]PX4t1!RR]PX4t2!RR]PX4t3!RR]PX4t4!R R!]PX4t5R9R"R#llt6!R$R%]PX4t7!R&R']PX4t8!R(R)]4t9!R*R+]PX4t:] !R,R-]44t;] !R.R/];44t<] !R0R1];44t=] !R2R37!R4R5];44t>] !R6R37!R7R8];44t?.R:Ot@R#);zPyTorch MarkupLM model.)CallableN)nn)BCEWithLogitsLossCrossEntropyLossMSELoss)initialization)ACT2FN)GradientCheckpointingLayer)BaseModelOutputBaseModelOutputWithPoolingMaskedLMOutputQuestionAnsweringModelOutputSequenceClassifierOutputTokenClassifierOutput)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)apply_chunking_to_forward)TransformersKwargsauto_docstringcan_return_tuplelogging)merge_with_config_defaults)capture_outputsMarkupLMConfigc@aa]tRt^-toRtV3RltRRltRtVtV;t #)XPathEmbeddingszConstruct the embeddings from xpath tags and subscripts. We drop tree-id in this version, as its info can be covered by xpath. c <\SV`4VPVn\P!VP VP,VP 4Vn\P!VP4Vn \P!4Vn \P!VP VP,^VP ,4Vn \P!^VP ,VP 4Vn\P!\!VP4Uu.uF.p\P"!VP$VP 4NK0 up4Vn\P!\!VP4Uu.uF.p\P"!VP(VP 4NK0 up4VnR#uupiuupi)N)super__init__ max_depthrLinearxpath_unit_hidden_size hidden_sizexpath_unitseq2_embeddingsDropouthidden_dropout_probdropoutReLU activationxpath_unitseq2_inner inner2emb ModuleListrange Embeddingmax_xpath_tag_unit_embeddingsxpath_tag_sub_embeddingsmax_xpath_subs_unit_embeddingsxpath_subs_sub_embeddings)selfconfig_ __class__s&& o/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/markuplm/modeling_markuplm.pyr!XPathEmbeddings.__init__3sd )))+63P3PSWSaSa3acicucu)v&zz&"<"<= '')$&IIf.K.Kdnn.\^_bhbtbt^t$u!1v'9'9#96;M;MN(* t~~. .A VAA6C`C`a. ) %*,t~~. .A VBBFDaDab. * &   s 4G0.4G5c .p.p\VP4FgpVPVPV,!VRRV3,44VPVPV,!VRRV3,44Ki \ P !VRR7p\ P !VRR7pW4,pVPVPVPVPV4444pV#)NNNdim) r/r"appendr2r4torchcatr-r)r+r,)r5xpath_tags_seqxpath_subs_seqxpath_tags_embeddingsxpath_subs_embeddingsixpath_embeddingss&&& r9forwardXPathEmbeddings.forwardMs " "t~~&A ! ( ()F)Fq)I.YZ\]_`Y`Ja)b c ! ( ()G)G)J>Z[]^`aZaKb)c d'!& *?R H % *?R H0H>>$,,ttG`G`aqGr7s*tu)r+r)r-r"r4r2r&r,)NN) __name__ __module__ __qualname____firstlineno____doc__r!rI__static_attributes____classdictcell__ __classcell__r8 __classdict__s@@r9rr-s  4  rKrcdaa]tRt^_toRtV3Rlt]R4t]RRl4tRRlt Rt Vt V;t #) MarkupLMEmbeddingszGConstruct the embeddings from word, position and token_type embeddings.c<\SV`4Wn\P!VP VP VPR7Vn\P!VPVP 4Vn VPVn \V4Vn \P!VPVP 4Vn\P !VP VP"R7Vn\P$!VP&4Vn\P*!\,P.!VP4P1R4RR7VnVPVn\P!VPVP VP4R7Vn R#)) padding_idxepsF) persistentNr?)r r!r6rr0 vocab_sizer% pad_token_idword_embeddingsmax_position_embeddingsposition_embeddingsr"rrHtype_vocab_sizetoken_type_embeddings LayerNormlayer_norm_epsr'r(r)BufferrAarangeexpand position_idsrYr5r6r8s&&r9r!MarkupLMEmbeddings.__init__bs/  !||F,=,=v?Q?Q_e_r_rs#%<<0N0NPVPbPb#c )) / 7%'\\&2H2H&J\J\%]"f&8&8f>S>STzz&"<"<= IIell63Q3Q&R&Y&YZa&botu!..#%<<  * *F,>,>DL\L\$  rKc VP4RRpV^,p\P!V^,W1,^,\PVPR7pVP ^4P V4#)z We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids. Args: inputs_embeds: torch.Tensor Returns: torch.Tensor Ndtypedevicer?)sizerArilongrq unsqueezerj) inputs_embedsrY input_shapesequence_lengthrks&& r9&create_position_ids_from_inputs_embeds9MarkupLMEmbeddings.create_position_ids_from_inputs_embedsxsn$((*3B/ %a.|| !O_:Q>ejjYfYmYm %%a(// <>*::;5::V\7N  !![[44uzzd;'4>>*::;5::V\8N)"66|D $ : :> J00P%;>SSVff ^^J/ \\*- rK) rfr6r)r"rYrcrkrerarH))NNNNNN) rLrMrNrOrPr! staticmethodrxrrIrQrRrSrTs@@r9rWrW_s?Q ,==" 8 811rKrWcDaa]tRt^toV3RltV3RlRltRtVtV;t#)MarkupLMSelfOutputc0<\SV`4\P!VPVP4Vn\P !VPVPR7Vn\P!VP4Vn R#rZN) r r!rr#r%denserfrgr'r(r)rls&&r9r!MarkupLMSelfOutput.__init__s` YYv1163E3EF f&8&8f>S>STzz&"<"<= rKch<V^8dQhRS[PRS[PRS[P/# hidden_states input_tensorreturnrATensor)formatrUs"r9 __annotate__MarkupLMSelfOutput.__annotate__1U\\RWR^R^rKcxVPV4pVPV4pVPW,4pV#Nrr)rfr5rrs&&&r9rIMarkupLMSelfOutput.forward5 =1  ]3 }'CD rKrfrr) rLrMrNrOr!rIrQrRrSrTs@@r9rr> rKrcDaa]tRt^toV3RltV3RlRltRtVtV;t#)MarkupLMIntermediatec"<\SV`4\P!VPVP 4Vn\VP\4'd\VP,Vn R#VPVn R#r) r r!rr#r%intermediate_sizer isinstance hidden_actstrrintermediate_act_fnrls&&r9r!MarkupLMIntermediate.__init__s` YYv1163K3KL f'' - -'-f.?.?'@D $'-'8'8D $rKcN<V^8dQhRS[PRS[P/#rrrr)rrUs"r9r!MarkupLMIntermediate.__annotate__s#U\\ellrKcJVPV4pVPV4pV#rrrr5rs&&r9rIMarkupLMIntermediate.forwards& =1 00? rKrrrTs@@r9rrs9rKrcDaa]tRt^toV3RltV3RlRltRtVtV;t#)MarkupLMOutputc0<\SV`4\P!VPVP 4Vn\P!VP VPR7Vn\P!VP4Vn R#r) r r!rr#rr%rrfrgr'r(r)rls&&r9r!MarkupLMOutput.__init__s` YYv779K9KL f&8&8f>S>STzz&"<"<= rKch<V^8dQhRS[PRS[PRS[P/#rr)rrUs"r9rMarkupLMOutput.__annotate__rrKcxVPV4pVPV4pVPW,4pV#rrrs&&&r9rIMarkupLMOutput.forwardrrKrrrTs@@r9rrrrKrcDaa]tRt^toV3RltV3RlRltRtVtV;t#)MarkupLMPoolerc<\SV`4\P!VPVP4Vn\P !4VnR#r)r r!rr#r%rTanhr+rls&&r9r!MarkupLMPooler.__init__s9 YYv1163E3EF '')rKcN<V^8dQhRS[PRS[P/#rr)rrUs"r9rMarkupLMPooler.__annotate__s#U\\ellrKc\VR,pVPV4pVPV4pV#)r<)r<r)rr+)r5rfirst_token_tensor pooled_outputs&& r9rIMarkupLMPooler.forwards2+40 #56  6 rK)r+rrrTs@@r9rrs$ rKrcDaa]tRtRtoV3RltV3RlRltRtVtV;t#)MarkupLMPredictionHeadTransformic<\SV`4\P!VPVP4Vn\ VP\4'd\VP,Vn MVPVn \P!VPVPR7Vn R#r) r r!rr#r%rrrrrtransform_act_fnrfrgrls&&r9r!(MarkupLMPredictionHeadTransform.__init__s~ YYv1163E3EF f'' - -$*6+<+<$=D !$*$5$5D !f&8&8f>S>STrKcN<V^8dQhRS[PRS[P/#rr)rrUs"r9r,MarkupLMPredictionHeadTransform.__annotate__s#U\\ellrKclVPV4pVPV4pVPV4pV#r)rrrfrs&&r9rI'MarkupLMPredictionHeadTransform.forwards4 =1 --m< }5 rK)rfrrrrTs@@r9rrsUrKrc8aa]tRtRtoV3RltRtRtVtV;t#)MarkupLMLMPredictionHeadi!c<\SV`4\V4Vn\P !VP VPRR7Vn\P!\P!VP44Vn R#)T)biasN) r r!r transformrr#r%r_decoder ParameterrArrrls&&r9r!!MarkupLMLMPredictionHead.__init__"s[ 8@yy!3!3V5F5FTR LLV->->!?@ rKcJVPV4pVPV4pV#r)rrrs&&r9rI MarkupLMLMPredictionHead.forward+s$}5  ]3 rK)rrrrrTs@@r9rr!sArKrcDaa]tRtRtoV3RltV3RlRltRtVtV;t#)MarkupLMOnlyMLMHeadi2cD<\SV`4\V4VnR#r)r r!r predictionsrls&&r9r!MarkupLMOnlyMLMHead.__init__3s 3F;rKcN<V^8dQhRS[PRS[P/#)rsequence_outputrr)rrUs"r9r MarkupLMOnlyMLMHead.__annotate__7s#!!u||! !rKc(VPV4pV#rr)r5rprediction_scoress&& r9rIMarkupLMOnlyMLMHead.forward7s ,,_=  rKrrrTs@@r9rr2s<!!!rKrcV^8dQhR\PR\PR\PR\PR\PR,R\R\/#) rmodulequerykeyvalueattention_maskNscalingr))rModulerArfloat)rs"r9rr=sg%% II% <<% % << % LL4' %  %%rKc \P!WP^^44V,pVe W,p\PP VR\P R7PVP4p\PPWVPR7p\P!W4p V P^^4P4p W3#)r)r>rp)ptrainingr?) rAmatmul transposer functionalsoftmaxfloat32torpr)r contiguous) rrrrrrr)kwargs attn_weights attn_outputs &&&&&&&, r9eager_attention_forwardr=s<<}}Q':;gEL!#4 ==((2U]](SVVW\WbWbcL==((6??([L,,|3K''1-88:K  $$rKcHaa]tRtRtoV3RltRV3RlRlltRtVtV;t#)MarkupLMSelfAttentioniTc\<\SV`4VPVP,^8wd9\ VR4'g'\ RVP RVP R24hWnVPVn\VPVP, 4VnVPVP,Vn \P!VPVP4Vn \P!VPVP4Vn \P!VPVP4Vn\P!VP 4VnVP VnVPR,VnR#)rembedding_sizezThe hidden size (z6) is not a multiple of the number of attention heads ()Ng)r r!r%num_attention_headshasattr ValueErrorr6r|attention_head_size all_head_sizerr#rrrr'attention_probs_dropout_probr)attention_dropoutrrls&&r9r!MarkupLMSelfAttention.__init__UsD     : : :a ?PVXhHiHi#F$6$6#78 445Q8   #)#=#= #&v'9'9F !JJL(M'+  rK)r6rGrFrrrTs@@r9rCrCs,    rKrCcvaa]tRtRtoRtR]R]/t]P!4V3Rl4t V3Rlt Rt Vt V;t#)MarkupLMPreTrainedModelimarkuplmr attentionsc <\SV`V4\V\4'd#\P !VP 4R#\V\4'db\P!VP\P!VPPR,4PR44R#R#)zInitialize the weightsNr?r])r _init_weightsrrinitzeros_rrWcopy_rkrArirrj)r5rr8s&&r9rT%MarkupLMPreTrainedModel._init_weightss~ f% f6 7 7 KK $  2 3 3 JJv**ELL9L9L9R9RSU9V,W,^,^_f,g h4rKc&<V^8dQh/S[;R&#)rr6r)rrUs"r9r$MarkupLMPreTrainedModel.__annotate__s rK)rLrMrNrObase_model_prefixr1r_can_record_outputsrAno_gradrT__annotate_func__rQrRrSrTs@@r9rPrPsC#+  ]]_iirKrPcvaa]tRtRtoRV3RlltRtRt]]] R V3RlRll444t Rt Vt V;t #) MarkupLMModelic <\SV`V4Wn\V4Vn\ V4VnV'd \V4MRVnVP4R#)z^ add_pooling_layer (bool, *optional*, defaults to `True`): Whether to add a pooling layer N) r r!r6rWrrCencoderrpooler post_init)r5r6add_pooling_layerr8s&&&r9r!MarkupLMModel.__init__sK   ,V4&v. 0AnV,t  rKc.VPP#rrra)r5s&r9get_input_embeddings"MarkupLMModel.get_input_embeddingss...rKc&WPnR#rri)r5rs&&r9set_input_embeddings"MarkupLMModel.set_input_embeddingss */'rKc^<V^8dQhRS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,R S[S[,R S[S[,/ #) rrNrCrDrrrkrurr)rA LongTensorrrrrr )rrUs"r9rMarkupLMModel.__annotate__ sK K ##d*K ((4/K ((4/ K ))D0 K ((4/ K &&-K ((4/K +,K  + +K rKc VeVe \R4hVe#VPW4VP4p M#VeVP4RR p M \R4hVe VPM VPp Vf\P !WR7pVf(\P !V \PV R7pVP^4P^4p V PVPR7p RV , R ,p VPVVVVVVR7p VP!V V 3/VBp V ^,pVPeVPV4MRp\VVR 7#) ao xpath_tags_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Tag IDs for each token in the input sequence, padded up to config.max_depth. xpath_subs_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Subscript IDs for each token in the input sequence, padded up to config.max_depth. Examples: ```python >>> from transformers import AutoProcessor, MarkupLMModel >>> processor = AutoProcessor.from_pretrained("microsoft/markuplm-base") >>> model = MarkupLMModel.from_pretrained("microsoft/markuplm-base") >>> html_string = " Page Title " >>> encoding = processor(html_string, return_tensors="pt") >>> outputs = model(**encoding) >>> last_hidden_states = outputs.last_hidden_state >>> list(last_hidden_states.shape) [1, 4, 768] ```NzDYou cannot specify both input_ids and inputs_embeds at the same timez5You have to specify either input_ids or inputs_embeds)rqro)rpg?)rrCrDrkrru)rL pooler_outputr?g)r %warn_if_padding_and_no_attention_maskrrrqrArrrsrtrrprrcrdr )r5rrCrDrrrkrurrvrqextended_attention_maskembedding_outputencoder_outputsrrs&&&&&&&&, r9rIMarkupLMModel.forward soJ  ]%>cd d  "  6 6y Q#..*K  &',,.s3KTU U%.%:!! @T@T  !"ZZ CN  !"[[EJJvVN"0":":1"="G"G"J"9"<"<4::"<"N#&)@#@H"L??))%)' + ,,  #   *!,8< 8O O4UY )-'  rK)r6rrcrd)T)NNNNNNN)rLrMrNrOr!rjrmrrrrIrQrRrSrTs@@r9rarasF /0 K K  K K rKrac\aa]tRtRtoV3Rlt]]RV3RlRll44tRtVt V;t #)MarkupLMForQuestionAnsweringiZc<\SV`V4VPVn\VRR7Vn\ P !VPVP4VnVP4R#F)rfN) r r! num_labelsrarQrr#r% qa_outputsrerls&&r9r!%MarkupLMForQuestionAnswering.__init__]sU   ++%fF ))F$6$68I8IJ rKc<V^8dQhRS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,R S[PR,R S[PR,R S[S[,R S[S[P,S[,/ #) rrNrCrDrrrkrustart_positions end_positionsrr)rArrrrr )rrUs"r9r)MarkupLMForQuestionAnswering.__annotate__isT T <<$&T  t+T  t+ T  t+ T  t+ T llT)T ||d*T ,T ||d*T +,T  u|| ; ;T rKc  VP!V3RVRVRVRVRVRV/V Bp V ^,p VPV 4p V P^R R7wrVPR 4P 4pVPR 4P 4pRpVeV e\ VP 44^8dVPR 4p\ V P 44^8dV PR 4p VP ^4pVP^V4V P^V4\VR 7pV!W4pV!W4pVV,^, p\VVVV PV PR 7#) a xpath_tags_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Tag IDs for each token in the input sequence, padded up to config.max_depth. xpath_subs_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Subscript IDs for each token in the input sequence, padded up to config.max_depth. Examples: ```python >>> from transformers import AutoProcessor, MarkupLMForQuestionAnswering >>> import torch >>> processor = AutoProcessor.from_pretrained("microsoft/markuplm-base-finetuned-websrc") >>> model = MarkupLMForQuestionAnswering.from_pretrained("microsoft/markuplm-base-finetuned-websrc") >>> html_string = " My name is Niels " >>> question = "What's his name?" >>> encoding = processor(html_string, questions=question, return_tensors="pt") >>> with torch.no_grad(): ... outputs = model(**encoding) >>> answer_start_index = outputs.start_logits.argmax() >>> answer_end_index = outputs.end_logits.argmax() >>> predict_answer_tokens = encoding.input_ids[0, answer_start_index : answer_end_index + 1] >>> processor.decode(predict_answer_tokens).strip() 'Niels' ```rCrDrrrkrur=N) ignore_index)loss start_logits end_logitsrrRr?) rQr~splitsqueezerlenrrclamp_rr rrR)r5rrCrDrrrkrurrroutputsrlogitsrr total_loss ignored_indexloss_fct start_lossend_losss&&&&&&&&&&, r9rI$MarkupLMForQuestionAnswering.forwardgsZ--   )  *  *  *  &  (    "!*1#)<<r<#: #++B/::< ''+668   &=+D?'')*Q."1"9"9""==%%'(1, - 5 5b 9 (--a0M  " "1m 4  M 2']CH!,@J :H$x/14J+%!!//))   rK)rQr}r~) NNNNNNNNN rLrMrNrOr!rrrIrQrRrSrTs@@r9rzrzZs4T T T T rKrzzC MarkupLM Model with a `token_classification` head on top. ) custom_introc\aa]tRtRtoV3Rlt]]RV3RlRll44tRtVt V;t #)MarkupLMForTokenClassificationicn<\SV`V4VPVn\VRR7VnVP e VP M VP p\P!V4Vn \P!VPVP4Vn VP4R#r|)r r!r}rarQclassifier_dropoutr(rr'r)r#r% classifierrer5r6rr8s&& r9r!'MarkupLMForTokenClassification.__init__s   ++%fF )/)B)B)NF % %TZTnTn zz"45 ))F$6$68I8IJ rKc<V^8dQhRS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,R S[PR,R S[S[,R S[S[P,S[,/ # rrNrCrDrrrkrulabelsrr)rArrrrr )rrUs"r9r+MarkupLMForTokenClassification.__annotate__sD D <<$&D  t+D  t+ D  t+ D  t+ D llT)D ||d*D  t#D +,D  u|| ~ -D rKc  RVP!V3RVRVRVRVRVRV/V Bp V ^,p VPV 4p Rp VeG\4pV!V PR VPP 4VPR 44p \ V V V PV PR7#) a xpath_tags_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Tag IDs for each token in the input sequence, padded up to config.max_depth. xpath_subs_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Subscript IDs for each token in the input sequence, padded up to config.max_depth. labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`. Examples: ```python >>> from transformers import AutoProcessor, AutoModelForTokenClassification >>> import torch >>> processor = AutoProcessor.from_pretrained("microsoft/markuplm-base") >>> processor.parse_html = False >>> model = AutoModelForTokenClassification.from_pretrained("microsoft/markuplm-base", num_labels=7) >>> nodes = ["hello", "world"] >>> xpaths = ["/html/body/div/li[1]/div/span", "/html/body/div/li[1]/div/span"] >>> node_labels = [1, 2] >>> encoding = processor(nodes=nodes, xpaths=xpaths, node_labels=node_labels, return_tensors="pt") >>> with torch.no_grad(): ... outputs = model(**encoding) >>> loss = outputs.loss >>> logits = outputs.logits ```rCrDrrrkruNrrrrRr?) rQrrrr6r}rrrR)r5rrCrDrrrkrurrrrrrrs&&&&&&&&&, r9rI&MarkupLMForTokenClassification.forwardsV--   )  *  *  *  &  (    "!* OOO<  ')H!&&r4;;+A+AB BD %$!//))   rK)rr)rQr}NNNNNNNNrrTs@@r9rrs4 D D D D rKrz MarkupLM Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled output) e.g. for GLUE tasks. c\aa]tRtRtoV3Rlt]]RV3RlRll44tRtVt V;t #)!MarkupLMForSequenceClassificationicv<\SV`V4VPVnWn\ V4VnVP e VP M VPp\P!V4Vn \P!VPVP4Vn VP4R#r)r r!r}r6rarQrr(rr'r)r#r%rrers&& r9r!*MarkupLMForSequenceClassification.__init__&s   ++ %f- )/)B)B)NF % %TZTnTn zz"45 ))F$6$68I8IJ rKc<V^8dQhRS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,RS[PR,R S[PR,R S[S[,R S[S[P,S[,/ #r)rArrrrr)rrUs"r9r.MarkupLMForSequenceClassification.__annotate__7sT T <<$&T  t+T  t+ T  t+ T  t+ T llT)T ||d*T  t#T +,T  u|| 7 7T rKc  VP!V3RVRVRVRVRVRV/V Bp V ^,p VPV 4p VPV 4p Rp VEeVPPfVP ^8XdRVPnMrVP ^8dQVP \P8Xg VP \P8XdR VPnMR VPnVPPR8XdK\4pVP ^8Xd'V!V P4VP44p MV!W4p MVPPR 8Xd>\4pV!V PR VP 4VPR 44p M-VPPR 8Xd\4pV!W4p \V V V P V P"R 7#) a  xpath_tags_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Tag IDs for each token in the input sequence, padded up to config.max_depth. xpath_subs_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Subscript IDs for each token in the input sequence, padded up to config.max_depth. labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If `config.num_labels > 1` a classification loss is computed (Cross-Entropy). Examples: ```python >>> from transformers import AutoProcessor, AutoModelForSequenceClassification >>> import torch >>> processor = AutoProcessor.from_pretrained("microsoft/markuplm-base") >>> model = AutoModelForSequenceClassification.from_pretrained("microsoft/markuplm-base", num_labels=7) >>> html_string = " Page Title " >>> encoding = processor(html_string, return_tensors="pt") >>> with torch.no_grad(): ... outputs = model(**encoding) >>> loss = outputs.loss >>> logits = outputs.logits ```rCrDrrrkruN regressionsingle_label_classificationmulti_label_classificationrr?)rQr)rr6 problem_typer}rprArsr|rrrrrrrrR)r5rrCrDrrrkrurrrrrrrs&&&&&&&&&, r9rI)MarkupLMForSequenceClassification.forward5sT--   )  *  *  *  &  (       ]3 /  {{''/??a'/;DKK,__q(fllejj.HFLL\a\e\eLe/LDKK,/KDKK,{{''<7"9??a'#FNN$4fnn6FGD#F3D))-JJ+- B @&++b/R))-II,./'!//))   rK)rr6r)rQr}rrrTs@@r9rrs4 T T T T rKr)r)rzrrrarP)ArPcollections.abcrrArtorch.nnrrrrrU activationsrmodeling_layersr modeling_outputsr r r r rrmodeling_utilsrrprocessing_utilsr pytorch_utilsrutilsrrrr utils.genericrutils.output_capturingrconfiguration_markuplmr get_loggerrLloggerrrrWrrrrrrrrrr%r1rCrPrarzrr__all__r[rKr9rs$ AA&!9G&6RR752   H %/ bii/ dnnd   299  RYY  RYY  bii$ ryy "!"))!%.3)BII3)n ..D bii 2ioii$f +f f Rb #:b b J V %<V  V r g (?g  g T rK