UhdZddlZddlZddlmZmZmZddlZddlZddlm Z ddl m Z m Z m Z ddlmZddlmZmZmZmZmZmZdd lmZmZmZmZdd lmZmZd d lm Z ejBe"Z#Gd de jHZ%d7dZ&Gdde jHZ'Gdde jHZ(Gdde jHZ)Gdde jHZ*Gdde jHZ+Gdde jHZ,Gdde jHZ-Gdde jHZ.Gd d!e jHZ/d"e/iZ0Gd#d$e jHZ1Gd%d&e jHZ2Gd'd(e jHZ3eGd)d*eZ4eGd+d,e4Z5eGd-d.e4Z6ed/0Gd1d2e4Z7ed30Gd4d5e4Z8gd6Z9y)8zPyTorch MarkupLM model.N)OptionalTupleUnion)nn)BCEWithLogitsLossCrossEntropyLossMSELoss)ACT2FN))BaseModelOutputWithPastAndCrossAttentions,BaseModelOutputWithPoolingAndCrossAttentionsMaskedLMOutputQuestionAnsweringModelOutputSequenceClassifierOutputTokenClassifierOutput)PreTrainedModelapply_chunking_to_forward find_pruneable_heads_and_indicesprune_linear_layer)auto_docstringlogging)MarkupLMConfigc*eZdZdZfdZddZxZS)XPathEmbeddingszConstruct the embeddings from xpath tags and subscripts. We drop tree-id in this version, as its info can be covered by xpath. c tt| |j|_t j |j |jz|j|_t j|j|_ t j|_ t j |j |jzd|jz|_t j d|jz|j|_t j t#|jDcgc],}t j$|j&|j .c}|_t j t#|jDcgc],}t j$|j*|j .c}|_ycc}wcc}w)N)superr__init__ max_depthrLinearxpath_unit_hidden_size hidden_sizexpath_unitseq2_embeddingsDropouthidden_dropout_probdropoutReLU activationxpath_unitseq2_inner inner2emb ModuleListrange Embeddingmax_xpath_tag_unit_embeddingsxpath_tag_sub_embeddingsmax_xpath_subs_unit_embeddingsxpath_subs_sub_embeddingsselfconfig_ __class__s /var/www/catia.catastroantioquia-mas.com/valormas/lib/python3.12/site-packages/transformers/models/markuplm/modeling_markuplm.pyrzXPathEmbeddings.__init__6s\ ot-/)))+63P3PSWSaSa3acicucu)v&zz&"<"<= '')$&IIf.K.Kdnn.\^_bhbtbt^t$u!1v'9'9#96;M;MN(* t~~.  VAA6C`C`a ) %*,t~~.  VBBFDaDab * &   s 51G1Gc g}g}t|jD]^}|j|j||dddd|f|j|j||dddd|f`t j |d}t j |d}||z}|j|j|j|j|}|S)Ndim) r-r appendr0r2torchcatr+r'r)r*)r4xpath_tags_seqxpath_subs_seqxpath_tags_embeddingsxpath_subs_embeddingsixpath_embeddingss r8forwardzXPathEmbeddings.forwardPs " "t~~& eA ! ( ()I)F)Fq)I.YZ\]_`Y`Ja)b c ! ( ()J)G)G)J>Z[]^`aZaKb)c d e!& *?R H % *?R H03HH>>$,,ttG`G`aqGr7s*tu)NN)__name__ __module__ __qualname____doc__rrF __classcell__r7s@r8rr0s  4 rGrc|j|j}tj|dj ||z|z}|j |zS)a Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols are ignored. This is modified from fairseq's `utils.make_positions`. Args: x: torch.Tensor x: Returns: torch.Tensor rr;)neintr>cumsumtype_aslong) input_ids padding_idxpast_key_values_lengthmaskincremental_indicess r8"create_position_ids_from_input_idsrYcsW << $ ( ( *D <<!4<eZdZdZfdZdZ ddZxZS)MarkupLMEmbeddingszGConstruct the embeddings from word, position and token_type embeddings.cltt| ||_t j |j |j|j|_ t j |j|j|_ |j|_ t||_t j |j|j|_t j"|j|j$|_t j&|j(|_|j-dt/j0|jj3dd|j|_t j |j|j|j4|_ y)N)rUeps position_ids)rr:F) persistent)rr[rr5rr. vocab_sizer# pad_token_idword_embeddingsmax_position_embeddingsposition_embeddingsr rrEtype_vocab_sizetoken_type_embeddings LayerNormlayer_norm_epsr%r&r'register_bufferr>arangeexpandrUr4r5r7s r8rzMarkupLMEmbeddings.__init__vs=  $02 !||F,=,=v?Q?Q_e_r_rs#%<<0N0NPVPbPb#c )) / 7%'\\&2H2H&J\J\%]"f&8&8f>S>STzz&"<"<=   ELL)G)GHOOPWXej  "..#%<<  * *F,>,>DL\L\$  rGc|jdd}|d}tj|jdz||jzdztj|j }|j dj|S)z We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids. Args: inputs_embeds: torch.Tensor Returns: torch.Tensor Nr:rdtypedevicer)sizer>rkrUrSrq unsqueezerl)r4 inputs_embeds input_shapesequence_lengthr_s r8&create_position_ids_from_inputs_embedsz9MarkupLMEmbeddings.create_position_ids_from_inputs_embedss$((*3B/ %a.||   q /D4D4D"Dq"HPUPZPZcpcwcw %%a(// <zerosrSrcr5 tag_pad_idonestuplelistr subs_pad_idrergrErhr')r4rTr@rAtoken_type_idsr_rtrVrurqwords_embeddingsrergrE embeddingss r8rFzMarkupLMEmbeddings.forwards  #..*K',,.s3K%.%:!! @T@T  $A)TM]M]_uv #JJ=Y  !"[[EJJvVN   00;M  !![[33ejjd;'4>>*::;5::V\7N  !![[44uzzd;'4>>*::;5::V\8N)"66|D $ : :> J00P%(;;>SSVff ^^J/ \\*- rG)NNNNNNr)rHrIrJrKrrwrFrLrMs@r8r[r[ss,Q 2=& 2rGr[cneZdZfdZdej dej dej fdZxZS)MarkupLMSelfOutputc(t|tj|j|j|_tj |j|j|_tj|j|_ yNr]) rrrr!r#denserhrir%r&r'rms r8rzMarkupLMSelfOutput.__init__s` YYv1163E3EF f&8&8f>S>STzz&"<"<= rG hidden_states input_tensorreturncr|j|}|j|}|j||z}|SNrr'rhr4rrs r8rFzMarkupLMSelfOutput.forward7 =1  ]3 }|'CD rGrHrIrJrr>TensorrFrLrMs@r8rr1> U\\RWR^R^rGrcVeZdZfdZdej dej fdZxZS)MarkupLMIntermediatect|tj|j|j |_t|jtrt|j|_ y|j|_ yr) rrrr!r#intermediate_sizer isinstance hidden_actstrr intermediate_act_fnrms r8rzMarkupLMIntermediate.__init__s] YYv1163K3KL f'' -'-f.?.?'@D $'-'8'8D $rGrrcJ|j|}|j|}|Sr)rrr4rs r8rFzMarkupLMIntermediate.forwards& =1 00? rGrrMs@r8rrs#9U\\ellrGrcneZdZfdZdej dej dej fdZxZS)MarkupLMOutputc(t|tj|j|j |_tj|j |j|_tj|j|_ yr) rrrr!rr#rrhrir%r&r'rms r8rzMarkupLMOutput.__init__s` YYv779K9KL f&8&8f>S>STzz&"<"<= rGrrrcr|j|}|j|}|j||z}|Srrrs r8rFzMarkupLMOutput.forwardrrGrrMs@r8rrrrGrcVeZdZfdZdej dej fdZxZS)MarkupLMPoolerct|tj|j|j|_tj |_yr)rrrr!r#rTanhr)rms r8rzMarkupLMPooler.__init__s9 YYv1163E3EF '')rGrrc\|dddf}|j|}|j|}|S)Nr)rr))r4rfirst_token_tensor pooled_outputs r8rFzMarkupLMPooler.forward s6+1a40 #56  6 rGrrMs@r8rrs#$ U\\ellrGrcVeZdZfdZdej dej fdZxZS)MarkupLMPredictionHeadTransformcht|tj|j|j|_t |jtrt|j|_ n|j|_ tj|j|j|_ yr) rrrr!r#rrrrr transform_act_fnrhrirms r8rz(MarkupLMPredictionHeadTransform.__init__s{ YYv1163E3EF f'' -$*6+<+<$=D !$*$5$5D !f&8&8f>S>STrGrrcl|j|}|j|}|j|}|Sr)rrrhrs r8rFz'MarkupLMPredictionHeadTransform.forwards4 =1 --m< }5 rGrrMs@r8rrs$UU\\ellrGrc*eZdZfdZdZdZxZS)MarkupLMLMPredictionHeadcHt|t||_t j |j |jd|_t jtj|j|_ |j|j_ y)NF)bias) rrr transformrr!r#radecoder Parameterr>ryrrms r8rz!MarkupLMLMPredictionHead.__init__'sm 8@yy!3!3V5F5FUS LLV->->!?@ !II rGc:|j|j_yr)rrr4s r8 _tie_weightsz%MarkupLMLMPredictionHead._tie_weights4s II rGcJ|j|}|j|}|Sr)rrrs r8rFz MarkupLMLMPredictionHead.forward7s$}5  ]3 rG)rHrIrJrrrFrLrMs@r8rr&s &&rGrcVeZdZfdZdej dej fdZxZS)MarkupLMOnlyMLMHeadcBt|t||_yr)rrr predictionsrms r8rzMarkupLMOnlyMLMHead.__init__?s 3F;rGsequence_outputrc(|j|}|Sr)r)r4rprediction_scoress r8rFzMarkupLMOnlyMLMHead.forwardCs ,,_=  rGrrMs@r8rr>s#<!u||! !rGrcPeZdZd fd Zdej dej fdZ ddej deejdeejdeejd eejd ee e ejd ee de ej fd Z xZ S)MarkupLMSelfAttentionct||j|jzdk7r2t |ds&t d|jd|jd|j|_t |j|jz |_|j|jz|_tj|j|j|_ tj|j|j|_ tj|j|j|_ tj|j|_|xs t#|dd|_|j$dk(s|j$d k(rF|j&|_tj(d |j&zd z |j|_|j,|_y) Nrembedding_sizezThe hidden size (z6) is not a multiple of the number of attention heads ()position_embedding_typeabsolute relative_keyrelative_key_queryr)rrr#num_attention_headshasattr ValueErrorrPattention_head_size all_head_sizerr!querykeyvaluer%attention_probs_dropout_probr'getattrrrdr.distance_embedding is_decoderr4r5rr7s r8rzMarkupLMSelfAttention.__init__Js     : : :a ?PVXhHi#F$6$6#78 445Q8  $*#=#= #&v'9'9F( ' -zC $  ' '> 9T=Y=Y]q=q+1+I+ID (&(ll1v7U7U3UXY3Y[_[s[s&tD # ++rGxrc|jdd|j|jfz}|j|}|j ddddS)Nr:rrrr )rrrrviewpermute)r4r new_x_shapes r8transpose_for_scoresz*MarkupLMSelfAttention.transpose_for_scoresdsLffhsmt'?'?AYAY&ZZ FF; yyAq!$$rGrattention_mask head_maskencoder_hidden_statesencoder_attention_maskpast_key_valueoutput_attentionsc$|j|}|du} | r||d} |d} |}n| rC|j|j|} |j|j|} |}n|y|j|j|} |j|j|} t j |d| gd} t j |d| gd} n@|j|j|} |j|j|} |j|} |du} |j r| | f}t j| | jdd}|jdk(s|jdk(r| jd| jd}}| rDt j|dz tj|j jdd}n@t j|tj|j jdd}t j|tj|j jdd}||z }|j!||j"zdz }|j%| j& }|jdk(rt j(d | |}||z}nE|jdk(r6t j(d | |}t j(d | |}||z|z}|t+j,|j.z }|||z}t0j2j5|d}|j7|}|||z}t j|| }|j9dddd j;}|j=dd|j>fz}|j|}|r||fn|f}|j r||fz}|S)Nrrrr;r:rrrorpzbhld,lrd->bhlrzbhrd,lrd->bhlrr ) rrrrr>r?rmatmul transposershapetensorrSrqrrkrrdtorpeinsummathsqrtrr functionalsoftmaxr'r contiguousrrr)r4rrrrrrrmixed_query_layeris_cross_attention key_layer value_layer query_layer use_cacheattention_scores query_length key_lengthposition_ids_lposition_ids_rdistancepositional_embeddingrelative_position_scoresrelative_position_scores_queryrelative_position_scores_keyattention_probs context_layernew_context_layer_shapeoutputss r8rFzMarkupLMSelfAttention.forwardis!JJ}5 3$> ."<&q)I(+K3N 11$((;P2QRI33DJJ?T4UVK3N  '11$((=2IJI33DJJ}4MNK >!#4i"@aHI))^A%6 $D!LK11$((=2IJI33DJJ}4MNK//0AB "$. ??(5N!<< Y5H5HR5PQ  ' '> 9T=Y=Y]q=q'2'8'8';Y__Q=O*L!&j1nEJJWdWkWk!l!q!q""'l%**UbUiUi!j!o!oprtu!v"\\*EJJ}OcOcdiijkmopN%6H#'#:#:8dFbFb;bef;f#g #7#:#:ARAR#:#S ++~=+0<<8H+Wk+l(#36N#N --1EE16>NP[]q1r./4||S"S%**+BC 6G=/2mM] ?? 11GrGrNNNNNF) rHrIrJrr>rrr FloatTensorrboolrFrLrMs@r8rrIs,4%ell%u||%7;15=A>BDH,1c||c!!2!23cE--. c ((9(9: c !)):): ; c!uU->->'?!@Ac$D>c u|| crGreagerceZdZd fd ZdZ d dej deejdeejdeejdeejdee e ejd ee d e ej fd Z xZ S)MarkupLMAttentionct|t|j|||_t ||_t|_y)Nr) rrMARKUPLM_SELF_ATTENTION_CLASSES_attn_implementationr4routputset pruned_headsrs r8rzMarkupLMAttention.__init__sC 3F4O4OP ,C  )0 ErGc>t|dk(ryt||jj|jj|j \}}t |jj||j_t |jj||j_t |jj||j_ t |jj|d|j_ |jjt|z |j_|jj|jjz|j_ |j j||_y)Nrrr;)lenrr4rrrrrrrr rrunion)r4headsindexs r8 prune_headszMarkupLMAttention.prune_headss u:? 7 49900$))2O2OQUQbQb u -TYY__eD *499==%@ ,TYY__eD .t{{/@/@%QO )- (E(EE (R %"&))"?"?$))B_B_"_  --33E:rGrrrrrrrrc p|j|||||||}|j|d|} | f|ddz} | S)Nrr)r4r ) r4rrrrrrr self_outputsattention_outputrs r8rFzMarkupLMAttention.forwardsWyy    ! "    ;;|A F#% QR(88rGrr) rHrIrJrrr>rrrrrrFrLrMs@r8rrs";*7;15=A>BDH,1||!!2!23E--.  ((9(9:  !)):): ; !uU->->'?!@A$D> u|| rGrceZdZfdZ d dej deejdeejdeejdeejdeeeejdee d eej fd Z d Z xZ S) MarkupLMLayercft||j|_d|_t ||_|j |_|j|_|jr,|j st|dt |d|_ t||_ t||_ y)Nrz> should be used as a decoder model if cross attention is addedrr )rrchunk_size_feed_forward seq_len_dimr attentionradd_cross_attentionrcrossattentionr intermediaterr rms r8rzMarkupLMLayer.__init__ s '-'E'E$*62 ++#)#=#=  # #?? D6)g!hii"3FT^"_D 08$V, rGrrrrrrrrc ||ddnd}|j|||||} | d} |jr | dd} | d} n| dd} d} |jrT|Rt|dstd|d||d dnd}|j | ||||||}|d} | |ddz} |d}  | z} t |j |j|j| }|f| z} |jr|  fz} | S) Nr)rrrrr:r z'If `encoder_hidden_states` are passed, z` has to be instantiated with cross-attention layers by setting `config.add_cross_attention=True`r) rrrrr rfeed_forward_chunkrr)r4rrrrrrrself_attn_past_key_valueself_attention_outputsrrpresent_key_valuecross_attn_present_key_valuecross_attn_past_key_valuecross_attention_outputs layer_outputs r8rFzMarkupLMLayer.forwards}:H9S>"1#5Y] !%   /3 "0" 2!4 ??,Qr2G 6r : ,QR0G'+$ ??4@4!12 =dVDDD @N?Yrs(;_c %&*&9&9 %&)!' # 7q9  7" ==G,C2+F ( 14P P 0  # #T%A%A4CSCSUe  /G+ ??!2 44GrGcL|j|}|j||}|Sr)r!r )r4rintermediate_outputr*s r8r#z MarkupLMLayer.feed_forward_chunkYs,"//0@A{{#68HI rGr) rHrIrJrr>rrrrrrFr#rLrMs@r8rr s -"7;15=A>BDH,1?||?!!2!23?E--. ? ((9(9: ? !)):): ; ?!uU->->'?!@A?$D>? u|| ?BrGrcDeZdZfdZ ddej deejdeejdeejdeejdeeeejdee d ee d ee d ee d e eej e ffd Z xZ S)MarkupLMEncoderct|||_tjt |j Dcgc] }t|c}|_d|_ ycc}w)NF) rrr5rr,r-num_hidden_layersrlayergradient_checkpointingr3s r8rzMarkupLMEncoder.__init__asN  ]]5IaIaCb#caM&$9#cd &+#$dsA#rrrrrpast_key_valuesrroutput_hidden_states return_dictrc | rdnd} |rdnd} |r|jjrdnd} |jr%|jr|rtj dd}|rdnd}t |jD]\}}| r| |fz} |||nd}|||nd}|jr/|jr#|j|j|||||||}n||||||||}|d}|r ||dfz }|s|| |dfz} |jjs| |dfz} | r| |fz} | std||| | | fDSt||| | | S) NzZ`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`...Frr:rrc3$K|]}|| ywrr7).0vs r8 z*MarkupLMEncoder.forward..s = s)last_hidden_stater3r attentionscross_attentions) r5rr2traininglogger warning_once enumerater1_gradient_checkpointing_func__call__r|r )r4rrrrrr3rrr4r5all_hidden_statesall_self_attentionsall_cross_attentionsnext_decoder_cacherD layer_modulelayer_head_maskr layer_outputss r8rFzMarkupLMEncoder.forwardgs#7BD$5b4%64;;;Z;Zr`d  & &4==##p" #,R$(4# VOA|#$58H$H!.7.CilO3B3N_Q/TXN**t}} $ A A ))!"#)*"% ! !-!"#)*"%! *!,M"}R'8&::" &9]1=Mrrrrrrr rFrLrMs@r8r.r.`s ,7;15=A>BEI$(,1/4&*S ||S !!2!23S E--. S ((9(9: S !)):): ; S "%e.?.?(@"ABS D>S $D>S 'tnS d^S  uU\\"$MM NS rGr.c`eZdZeZdZdZedee e e jfffd Z xZS)MarkupLMPreTrainedModelmarkuplmclt|tjrm|jjj d|j j|j%|jjjyyt|tjrz|jjj d|j j|j2|jj|jjyyt|tjrJ|jjj|jjjdyt|tr%|jjjyy)zInitialize the weightsg)meanstdN?)rrr!weightdatanormal_r5initializer_rangerzero_r.rUrhfill_r)r4modules r8 _init_weightsz%MarkupLMPreTrainedModel._init_weightss' fbii ( MM   & &CT[[5R5R & S{{&   &&('  - MM   & &CT[[5R5R & S!!- ""6#5#56<<>.  - KK   " " $ MM   $ $S )  8 9 KK   " " $:rGpretrained_model_name_or_pathc2tt| |g|i|Sr)rrMfrom_pretrained)clsr[ model_argskwargsr7s r8r]z'MarkupLMPreTrainedModel.from_pretraineds+,cB ) ,6 :@  rG)rHrIrJr config_classbase_model_prefixrZ classmethodrrrosPathLiker]rLrMs@r8rMrMsD!L"%$ HU3PRP[P[K[E\<]  rGrMc~eZdZdfd ZdZdZdZe ddee jdee jdee jdee jd ee jd ee jd ee jd ee jd ee dee dee de eeffdZdZxZS) MarkupLMModelct||||_t||_t ||_|r t|nd|_|jy)zv add_pooling_layer (bool, *optional*, defaults to `True`): Whether to add a pooling layer N) rrr5r[rr.encoderrpooler post_init)r4r5add_pooling_layerr7s r8rzMarkupLMModel.__init__sM   ,V4&v. 0AnV,t  rGc.|jjSrrrcrs r8get_input_embeddingsz"MarkupLMModel.get_input_embeddingss...rGc&||j_yrrn)r4rs r8set_input_embeddingsz"MarkupLMModel.set_input_embeddingss*/'rGc|jD]7\}}|jj|jj |9y)z Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTrainedModel N)itemsrir1rr)r4heads_to_pruner1rs r8 _prune_headszMarkupLMModel._prune_headssE +002 CLE5 LL  u % / / ; ;E B CrGrTr@rArrr_rrtrr4r5rc | | n|jj} | | n|jj} | | n|jj} | | t d|#|j |||j } n!||j dd} n t d| |jn |j} |tj| | }|&tj| tj| }|jdjd}|j|j }d |z d z}||jdk(rh|jd jd jdjd}|j!|jj"dddd}nB|jdk(r/|jdjdjd}|jt%|j'j }ndg|jj"z}|j)|||||| }|j+|||| | | }|d }|j,|j-|nd}| s ||f|ddzSt/|||j0|j2|j4S)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, 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 timer:z5You have to specify either input_ids or inputs_embeds)rqrorrrrRgr)rTr@rAr_rrt)rrr4r5)r< pooler_outputrr=r>)r5rr4use_return_dictr%warn_if_padding_and_no_attention_maskrrrqr>r{ryrSrsrrpr<rlr0next parametersrrirjr rr=r>)r4rTr@rArrr_rrtrr4r5rurqextended_attention_maskembedding_outputencoder_outputsrrs r8rFzMarkupLMModel.forwardsL2C1N-TXT_T_TqTq$8$D $++JjJj &1%cd d  "  6 6y. Q#..*K  &',,.s3KTU U%.%:!! @T@T  !"ZZ FCN  !"[[EJJvVN"0":":1"="G"G"J"9"<"<4::"<"N#&)@#@H"L  }}!#%//2<!>>I??))%)' + ,,  #/!5# ' *!,8< 8O O4UY #]3oab6II I;-')77&11,==   rGcJd}|D]}|tfd|Dfz }|S)Nr7c3tK|]/}|jdj|j1yw)rN) index_selectrrq)r9 past_statebeam_idxs r8r;z/MarkupLMModel._reorder_cache..ms.nU_j--aZ=N=N1OPns58)r|)r4r3rreordered_past layer_pasts ` r8_reorder_cachezMarkupLMModel._reorder_cacheis=) J ncmnn N rG)T) NNNNNNNNNNN)rHrIrJrrorqrurrr> LongTensorrrrrr rFrrLrMs@r8rgrgsH /0C1559596:59371559,0/3&*h E,,-h !!1!12h !!1!12 h !!2!23 h !!1!12 h u//0h E--.h   1 12h $D>h 'tnh d^h  uBB Ch h VrGrgc eZdZfdZe ddeejdeejdeejdeejdeejdeejdeejd eejd eejd eejd eed eedeede e eje ffdZ xZ S)MarkupLMForQuestionAnsweringct|||j|_t|d|_t j |j|j|_|jyNF)rl) rr num_labelsrgrNrr!r# qa_outputsrkrms r8rz%MarkupLMForQuestionAnswering.__init__usU   ++%fF ))F$6$68I8IJ rGrTr@rArrr_rrtstart_positions end_positionsrr4r5rc ,| | n|jj} |j||||||||| | |  }|d}|j|}|j dd\}}|j dj }|j dj }d}| | t| jdkDr| j d} t| jdkDr| j d} |jd}| jd|| jd|t|}||| }||| }||zdz }| s||f|ddz}||f|zS|St||||j|j S) ae xpath_tags_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Tag IDs for each token in the input sequence, padded up to config.max_depth. xpath_subs_seq (`torch.LongTensor` of shape `(batch_size, sequence_length, config.max_depth)`, *optional*): Subscript IDs for each token in the input sequence, padded up to config.max_depth. Examples: ```python >>> from transformers import AutoProcessor, MarkupLMForQuestionAnswering >>> import torch >>> processor = AutoProcessor.from_pretrained("microsoft/markuplm-base-finetuned-websrc") >>> model = MarkupLMForQuestionAnswering.from_pretrained("microsoft/markuplm-base-finetuned-websrc") >>> html_string = " My name is Niels " >>> question = "What's his name?" >>> encoding = processor(html_string, questions=question, return_tensors="pt") >>> with torch.no_grad(): ... outputs = model(**encoding) >>> answer_start_index = outputs.start_logits.argmax() >>> answer_end_index = outputs.end_logits.argmax() >>> predict_answer_tokens = encoding.input_ids[0, answer_start_index : answer_end_index + 1] >>> processor.decode(predict_answer_tokens).strip() 'Niels' ```N r@rArrr_rrtrr4r5rrr:r;) ignore_indexr)loss start_logits end_logitsrr=)r5rxrNrsplitsqueezerrrrclamp_rrrr=)r4rTr@rArrr_rrtrrrr4r5rrlogitsrr total_loss ignored_indexloss_fct start_lossend_lossr s r8rFz$MarkupLMForQuestionAnswering.forwards^&1%rrrrrrFrLrMs@r8rrrsY-115151515/3,0042604,0/3&*` ELL)` !.` !. ` !. ` !. ` u||,` ELL)`   -` "%,,/`   -` $D>` 'tn` d^`  uU\\"$@@ A` ` rGrzC MarkupLM Model with a `token_classification` head on top. ) custom_introceZdZfdZe ddeejdeejdeejdeejdeejdeejdeejd eejd eejd eed eed eede e eje ffdZ xZ S)MarkupLMForTokenClassificationcdt|||j|_t|d|_|j |j n |j }tj||_ tj|j|j|_ |jyr)rrrrgrNclassifier_dropoutr&rr%r'r!r# classifierrkr4r5rr7s r8rz'MarkupLMForTokenClassification.__init__s   ++%fF )/)B)B)NF % %TZTnTn zz"45 ))F$6$68I8IJ rGrTr@rArrr_rrtlabelsrr4r5rc | | n|jj} |j||||||||| | |  } | d}|j|}d}| Ft }||j d|jj | j d}| s|f| ddz}||f|zS|St||| j| jS)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 ```Nrrr:rrrrr=) r5rxrNrrrrrrr=)r4rTr@rArrr_rrtrrr4r5rrrrrr s r8rFz&MarkupLMForTokenClassification.forwardsZ&1%rrrrrrFrLrMs@r8rrsA -115151515/3,004)-,0/3&*P ELL)P !.P !. P !. P !. P u||,P ELL)P   -P &P $D>P 'tnP d^P  uU\\"N2 3P P rGrz 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. ceZdZfdZe ddeejdeejdeejdeejdeejdeejdeejd eejd eejd eed eed eede e eje ffdZ xZ S)!MarkupLMForSequenceClassificationcnt|||j|_||_t ||_|j |j n |j}tj||_ tj|j|j|_ |jyr)rrrr5rgrNrr&rr%r'r!r#rrkrs r8rz*MarkupLMForSequenceClassification.__init__Ts   ++ %f- )/)B)B)NF % %TZTnTn zz"45 ))F$6$68I8IJ rGrTr@rArrr_rrtrrr4r5rc D| | n|jj} |j||||||||| | |  } | d}|j|}|j |}d}| |jj |j dk(rd|j_nl|j dkDrL| jtjk(s| jtjk(rd|j_nd|j_|jj dk(rIt}|j dk(r&||j| j}n||| }n|jj dk(r=t}||jd|j | jd}n,|jj dk(rt}||| }| s|f| ddz}||f|zS|St!||| j"| j$ S) 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 ```Nrr regressionsingle_label_classificationmulti_label_classificationr:rr)r5rxrNr'r problem_typerrpr>rSrPr rrrrrrr=)r4rTr@rArrr_rrtrrr4r5rrrrrr s r8rFz)MarkupLMForSequenceClassification.forwardcsX&1%rrrrrrFrLrMs@r8rrLsB -115151515/3,004)-,0/3&*_ ELL)_ !._ !. _ !. _ !. _ u||,_ ELL)_   -_ &_ $D>_ 'tn_ d^_  uU\\"$<< =_ _ rGr)rrrrgrM)r):rKrrdtypingrrrr>torch.utils.checkpointrtorch.nnrrr activationsr modeling_outputsr r rrrrmodeling_utilsrrrrutilsrrconfiguration_markuplmr get_loggerrHr@ModulerrYr[rrrrrrrrr rrr.rMrgrrr__all__r7rGr8rs )) AA! -2   H %/ bii/ f 4 __F   299  RYY  RYY  bii$ryy0!"))!CBIICN "# 0 0hSBIISnZ biiZ z o  <R+RRjm #:m m ` a %<a  a H q (?q  q h rG