fThZSrSSKrSSKJrJrJr SSKrSSKrSSKJr SSK J r J r J r SSK Jr SSKJrJrJrJrJr SS KJr SS KJrJrJr SS KJrJr S S KJr \R@"\!5r""SS\RF5r$"SS\RF5r%"SS\RF5r&"SS\RF5r'"SS\RF5r("SS\RF5r)"SS\RF5r*"SS\RF5r+"SS\RF5r,"S S!\RF5r-\"S"S#\55r.\"S$S%\.55r/\"S&S'9"S(S)\.55r0\"S*S+\.55r1"S,S-\RF5r2\"S.S/\.55r3/S0Qr4g)1zPyTorch LiLT model.N)OptionalTupleUnion)nn)BCEWithLogitsLossCrossEntropyLossMSELoss)ACT2FN)BaseModelOutputBaseModelOutputWithPoolingQuestionAnsweringModelOutputSequenceClassifierOutputTokenClassifierOutput)PreTrainedModel)apply_chunking_to_forward find_pruneable_heads_and_indicesprune_linear_layer)auto_docstringlogging) LiltConfigcF^\rSrSrU4SjrSSjrSrSrSrU=r $)LiltTextEmbeddings*c.>[TU]5 [R"URUR UR S9Ul[R"URUR 5Ul [R"URUR 5Ul [R"UR URS9Ul [R"UR5UlUR#S[$R&"UR5R)S5SS9 [+USS5UlUR Ul[R"URUR UR.S9Ul g) N padding_idxeps position_ids)rF) persistentposition_embedding_typeabsolute)super__init__r Embedding vocab_size hidden_size pad_token_idword_embeddingsmax_position_embeddingsposition_embeddingstype_vocab_sizetoken_type_embeddings LayerNormlayer_norm_epsDropouthidden_dropout_probdropoutregister_buffertorcharangeexpandgetattrr$rselfconfig __class__s ^/var/www/auris/envauris/lib/python3.13/site-packages/transformers/models/lilt/modeling_lilt.pyr'LiltTextEmbeddings.__init__+s2 !||F,=,=v?Q?Q_e_r_rs#%<<0N0NPVPbPb#c %'\\&2H2H&J\J\%]"f&8&8f>S>STzz&"<"<=   ELL)G)GHOOPWXej  (/v7PR\']$"..#%<<  * *F,>,>DL\L\$  cDUcIUb5URXR5RUR5nOUR U5nUbUR 5nOUR 5SSnUc8[ R"U[ RURRS9nUcURU5nURU5nXF-nURS:XaURU5nXx- nURU5nURU5nXs4$)Nr"dtypedevicer%)"create_position_ids_from_input_idsrtorE&create_position_ids_from_inputs_embedssizer7zeroslongr!r,r0r$r.r1r5) r< input_idstoken_type_idsr! inputs_embeds input_shaper0 embeddingsr.s r?forwardLiltTextEmbeddings.forwardBs  $#FFyRbRbcff$$   $JJ=Y  #..*K',,.s3K  !"[[EJJtO`O`OgOghN   00;M $ : :> J":  ' ': 5"&":":<"H   -J^^J/ \\*- ''rAcURU5R5n[R"USS9R U5U-nUR 5U-$)z Args: 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`. x: torch.Tensor x: Returns: torch.Tensor rdim)neintr7cumsumtype_asrK)r<rLrmaskincremental_indicess r?rF5LiltTextEmbeddings.create_position_ids_from_input_idsfsP||K(,,.$||Da8@@F$N"'')K77rAcUR5SSnUSn[R"URS-X0R-S-[RUR S9nUR S5RU5$)z Args: We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.: inputs_embeds: torch.Tensor Returns: torch.Tensor Nr"rrCr)rIr7r8rrKrE unsqueezer9)r<rNrOsequence_lengthr!s r?rH9LiltTextEmbeddings.create_position_ids_from_inputs_embedsss~$((*3B/ %a.||   q /4D4D"Dq"HPUPZPZcpcwcw %%a(// <s@r?rr*s+ 2 "(H 8 = =rArc2^\rSrSrU4SjrSSjrSrU=r$)LiltLayoutEmbeddingsc>[TU]5 [R"URUR S-5Ul[R"URUR S-5Ul[R"URUR S-5Ul[R"URUR S-5Ul URUl [R"URUR UR-URS9Ul[R"UR UR UR-S9Ul[R""UR UR-UR$S9Ul[R&"UR(5Ulg)Nr) in_features out_featuresr)r&r'rr(max_2d_position_embeddingsr*x_position_embeddingsy_position_embeddingsh_position_embeddingsw_position_embeddingsr+rr-channel_shrink_ratiobox_position_embeddingsLinearbox_linear_embeddingsr1r2r3r4r5r;s r?r'LiltLayoutEmbeddings.__init__s^ &(\\&2S2SU[UgUgklUl%m"%'\\&2S2SU[UgUgklUl%m"%'\\&2S2SU[UgUgklUl%m"%'\\&2S2SU[UgUgklUl%m"!..')||  * *   &"="= =((( $ &(YY**9K9KvOjOj9j& "f&8&8F">|"L&A&[#&*nn5P&Q#&*ll3N&O#**3 cZ[ab b csA,D## D>- D99D>) r1rwrur5rrrrsrprq)NN)rarbrcrdr'rQrerfrgs@r?riris>*++rAricF^\rSrSrSU4SjjrSSjrSSjrSrU=r$) LiltSelfAttentionc>[TU]5 URUR-S:wa7[ US5(d&[ SURSURS35eURUl[ URUR- 5UlURUR-Ul[R"URUR5Ul [R"URUR5Ul [R"URUR5Ul [R"URUR-URUR-5Ul[R"URUR-URUR-5Ul[R"URUR-URUR-5Ul[R$"UR&5UlU=(d [+USS5UlUR,S:XdUR,S :XaGUR.Ul[R0"S UR.-S - UR5UlURUlg) Nrembedding_sizezThe hidden size (z6) is not a multiple of the number of attention heads ()r$r% relative_keyrelative_key_queryrzr)r&r'r*num_attention_headshasattr ValueErrorrWattention_head_size all_head_sizerrvquerykeyvaluert layout_query layout_key layout_valuer3attention_probs_dropout_probr5r:r$r-r(distance_embedding)r<r=r$r>s r?r'LiltSelfAttention.__init__sQ     : : :a ?PVXhHiHi#F$6$6#78 445Q8  $*#=#= #&v'9'9F( ' -zC $  ' '> 9T=Y=Y]q=q+1+I+ID (&(ll1v7U7U3UXY3Y[_[s[s&tD #$*$?$?!rAcUR5SSURURU-4-nUR"U6nUR SSSS5$)Nr"rrzrr )rIrrviewpermute)r<xr new_x_shapes r?transpose_for_scores&LiltSelfAttention.transpose_for_scoressRffhsmt'?'?AYAY]^A^&__ FFK yyAq!$$rAcURURU5URS9nURURU5URS9nURUR U5URS9nUR U5n URUR U55n URURU55n URU 5n [R"XRSS55n [R"XRSS55nURS:XdURS:XGaDUR5Sn[R"U[RURS9R!SS5n[R"U[RURS9R!SS5nUU- nUR#UUR$-S- 5nUR'U R(S9nURS:Xa[R*"S U U5nU U-n OHURS:Xa8[R*"S U U5n[R*"S U U5nU U-U-n U [,R."UR05- nU[,R."UR0UR-5- nUU-n UU-nUbX-n[2R4"SS 9"U5nUR7U5nUbUU-n[R"UU5nUR9S S SS5R;5nUR5SSUR<UR-4-nUR "U6nUbX-n [2R4"SS 9"U 5nUR7U5nUbUU-n[R"UU 5nUR9S S SS5R;5nUR5SSUR<4-nUR "U6nU(aUU4U4nU$UU44nU$)N)rr"rrrrC)rDzbhld,lrd->bhlrzbhrd,lrd->bhlrrTrrzr )rrrtrrrrrr7matmul transposer$rIr8rKrErrr-rGrDeinsummathsqrtrrSoftmaxr5r contiguousr)r< hidden_states layout_inputsattention_mask head_maskoutput_attentionslayout_value_layerlayout_key_layerlayout_query_layermixed_query_layer key_layer value_layer query_layerattention_scoreslayout_attention_scores seq_lengthposition_ids_lposition_ids_rdistancepositional_embeddingrelative_position_scoresrelative_position_scores_queryrelative_position_scores_keytmp_attention_scorestmp_layout_attention_scoreslayout_attention_probslayout_context_layernew_context_layer_shapeattention_probs context_layeroutputss r?rQLiltSelfAttention.forwardsR"66t7H7H7W[_[t[t6u44T__]5SW[WpWp4q!66t7H7H7W[_[t[t6u JJ}5--dhh}.EF // =0IJ //0AB  << 5H5HR5PQ"',,/AC]C]^`bdCe"f  ' '> 9T=Y=Y]q=q&++-a0J"\\*EJJ}OcOcdiijlnopN"\\*EJJ}OcOcdiijkmopN%6H#'#:#:8dFbFb;bef;f#g #7#:#:ARAR#:#S ++~=+0<<8H+Wk+l(#36N#N --1EE16>NP[]q1r./4||,,7  - 9O _kB %--aAq9DDF "/"4"4"6s";t?Q?Q>S"S%**,CD !1 2O D  !"679  rA)rrrtrr5rrrrr-rr$rrN)rNNF) rarbrcrdr'rrQrerfrgs@r?rrs&"@H% \\rArcz^\rSrSrU4SjrS\R S\R S\R 4SjrSrU=r $)LiltSelfOutputiEc(>[TU]5 [R"URUR5Ul[R "URURS9Ul[R"UR5Ul gNr) r&r'rrvr*denser1r2r3r4r5r;s r?r'LiltSelfOutput.__init__Fs` YYv1163E3EF f&8&8f>S>STzz&"<"<= rAr input_tensorreturncpURU5nURU5nURX-5nU$rrr5r1r<rrs r?rQLiltSelfOutput.forwardL5 =1  ]3 }'CD rAr1rr5 rarbrcrdr'r7TensorrQrerfrgs@r?rrE6> U\\RWR^R^rArc^\rSrSrS U4SjjrSrS S\RS\RS\\RS\\RS\\ S \ \R4 S jjr S r U=r$) LiltAttentioniSc>[TU]5 [XS9Ul[ U5Ul[ 5UlURnURUR-Ul[ U5Ul X1lg)N)r$) r&r'rr<routputset pruned_headsr*rt layout_output)r<r=r$ori_hidden_sizer>s r?r'LiltAttention.__init__Tsg %f^ $V, E ,,#//63N3NN+F3,rAc6[U5S:Xag[XRRURRUR 5up[ URRU5URl[ URRU5URl[ URRU5URl [ URRUSS9URl URR[U5- URlURRURR-URl UR RU5Ulg)NrrrT)lenrr<rrrrrrrrrrunion)r<headsindexs r? prune_headsLiltAttention.prune_heads`s  u:? 7 9900$))2O2OQUQbQb  -TYY__eD *499==%@ ,TYY__eD .t{{/@/@%QO )- (E(EE (R %"&))"?"?$))B_B_"_  --33E:rArrrrrrcURUUUUU5nURUSSU5nURUSSU5nXx44USS-n U $)Nrr)r<rr) r<rrrrr self_outputsattention_outputlayout_attention_outputrs r?rQLiltAttention.forwardrswyy        ;;|Aq'9=I"&"4"4\!_Q5G"W$>@  u|| rArcb^\rSrSrU4SjrS\R S\R 4SjrSrU=r $)LiltIntermediateic>[TU]5 [R"URUR 5Ul[UR[5(a[URUl gURUl gr) r&r'rrvr*intermediate_sizer isinstance hidden_actstrr intermediate_act_fnr;s r?r'LiltIntermediate.__init__s` YYv1163K3KL f'' - -'-f.?.?'@D $'-'8'8D $rArrcJURU5nURU5nU$rrr)r<rs r?rQLiltIntermediate.forwards& =1 00? rArrrgs@r?rrs(9U\\ellrArcz^\rSrSrU4SjrS\R S\R S\R 4SjrSrU=r $) LiltOutputic(>[TU]5 [R"URUR 5Ul[R"UR URS9Ul[R"UR5Ul gr) r&r'rrvrr*rr1r2r3r4r5r;s r?r'LiltOutput.__init__s` YYv779K9KL f&8&8f>S>STzz&"<"<= rArrrcpURU5nURU5nURX-5nU$rrrs r?rQLiltOutput.forwardrrArrrgs@r?rrrrArc^\rSrSrU4SjrS S\R S\R S\\RS\\RS\\ S\ \R 4 S jjr S r S r S rU=r$) LiltLayeric>[TU]5 URUlSUl[ U5Ul[ U5Ul[U5Ul URnURnURUR-Ul URUR-Ul [ U5Ul [U5UlX!l X1l g)Nr)r&r'chunk_size_feed_forward seq_len_dimr attentionr intermediaterrr*rrtlayout_intermediater)r<r=rori_intermediate_sizer>s r?r'LiltLayer.__init__s '-'E'E$&v.,V4 (  ,, & 8 8#//63N3NN#)#;#;v?Z?Z#Z #3F#; '/,#8 rArrrrrrcURUUUUUS9nUSSnUSSnUSSn [URURURU5n [UR URURU5n X44U -n U $)N)rrr)r rfeed_forward_chunkr r layout_feed_forward_chunk) r<rrrrrself_attention_outputsrrr layer_outputlayout_layer_outputs r?rQLiltLayer.forwards"&    / "0" 2!4Q7"8";A">(,0  # #T%A%A4CSCSUe 8  * *D,H,H$JZJZ\s !687BrAcJURU5nURX!5nU$r)r rr<rintermediate_outputrs r?rLiltLayer.feed_forward_chunks)"//0@A{{#6I rAcJURU5nURX!5nU$r)rrrs r?r#LiltLayer.layout_feed_forward_chunks+"667GH))*=P rA)r r r rrrr r)rarbrcrdr'r7rrrrrrQrrrerfrgs@r?rrs9*7;15,1 ||||!!2!23  E--.  $D>  u|| > rArc^\rSrSrU4SjrS S\R S\R S\\RS\\RS\\ S\\ S \\ S \ \ \R \ 44S jjr S rU=r$) LiltEncoderic>[TU]5 Xl[R"[ UR 5Vs/sHn[U5PM sn5UlSUl gs snf)NF) r&r'r=r ModuleListrangenum_hidden_layersrlayergradient_checkpointing)r<r=_r>s r?r'LiltEncoder.__init__sR  ]]uVE]E]?^#_?^!If$5?^#_` &+#$`sA&rrrrroutput_hidden_states return_dictrc U(aSOSnU(aSOSn [UR5HupU(aX4-nUbXJOSn UR(a2UR(a!UR U R UUUU U5n O U "UUUU U5n U SSnU SSnU(dMXS4-n M U(aX4-nU(d[ SUUU 455$[UUU S9$)Nrrc30# UH nUcMUv M g7frr+).0vs r? &LiltEncoder.forward..s"A  s )last_hidden_stater attentions) enumerater$r%training_gradient_checkpointing_func__call__tupler )r<rrrrrr(r)all_hidden_statesall_self_attentionsi layer_modulelayer_head_mask layer_outputss r?rQLiltEncoder.forwards&#7BD$5b4(4OA#$58H$H!.7.CilO**t}} $ A A ))!!"#% ! !-!!"#% ! *!,Q/M)!,Q/M  &91=M < 'tn< d^<  uU\\"O3 4< < rArcb^\rSrSrU4SjrS\R S\R 4SjrSrU=r $) LiltPooleri*c>[TU]5 [R"URUR5Ul[R "5Ulgr)r&r'rrvr*rTanh activationr;s r?r'LiltPooler.__init__+s9 YYv1163E3EF '')rArrc\USS2S4nURU5nURU5nU$Nr)rrC)r<rfirst_token_tensor pooled_outputs r?rQLiltPooler.forward0s6+1a40 #56  6 rA)rCrrrgs@r?r@r@*s($ U\\ellrAr@c*\rSrSr\rSrSr/rSr Sr g)LiltPreTrainedModeli9liltTc[U[R5(akURRR SUR RS9 URb%URRR5 gg[U[R5(axURRR SUR RS9 URb2URRURR5 gg[U[R5(aJURRR5 URRRS5 gg)zInitialize the weightsg)meanstdNg?)rrrvweightdatanormal_r=initializer_rangebiaszero_r(rr1fill_)r<modules r? _init_weights!LiltPreTrainedModel._init_weights@s fbii ( ( MM   & &CT[[5R5R & S{{&   &&('  - - MM   & &CT[[5R5R & S!!- ""6#5#56<<>.  - - KK   " " $ MM   $ $S ).rAr+N) rarbrcrdr config_classbase_model_prefixsupports_gradient_checkpointing_no_split_modulesrXrer+rAr?rKrK9sL&*#*rArKc^\rSrSrSU4SjjrSrSrSr\SS\ \ RS\ \ RS\ \ RS \ \ RS \ \ RS \ \ RS \ \ RS \ \ S\ \ S\ \ S\ \\ R\44Sjj5rSrU=r$) LiltModeliQc>[TU]U5 Xl[U5Ul[ U5Ul[U5UlU(a [U5OSUl UR5 g)z^ add_pooling_layer (bool, *optional*, defaults to `True`): Whether to add a pooling layer N) r&r'r=rrPrilayout_embeddingsrencoderr@pooler post_init)r<r=add_pooling_layerr>s r?r'LiltModel.__init__SsX   ,V4!5f!="6* ,=j(4  rAc.URR$rrPr,)r<s r?get_input_embeddingsLiltModel.get_input_embeddingsds...rAc$XRlgrrh)r<rs r?set_input_embeddingsLiltModel.set_input_embeddingsgs */'rAcUR5H7up#URRURR U5 M9 g)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)itemsrbr$r r)r<heads_to_pruner$rs r? _prune_headsLiltModel._prune_headsjs< +002LE LL  u % / / ; ;E B3rArLr}rrMr!rrNrr(r)rc UbUOURRnU bU OURRn U bU OURRn UbUb [ S5eUb"UR X5 UR 5n O"UbUR 5SSn O [ S5eU upUb URO URnUc'[R"U S-[RUS9nUc[R"X4US9nUcr[URS5(a3URRSS2SU 24nURX5nUnO$[R"U [RUS9nUR!X;5nUR#X`RR$5nURUUUUS 9unnUR'X%S 9nUR)UUUUUU U S 9nUS nUR*bUR+U5OSnU (d UU4US S-$[-UUUR.UR0S9$)aE bbox (`torch.LongTensor` of shape `(batch_size, sequence_length, 4)`, *optional*): Bounding boxes of each input sequence tokens. Selected in the range `[0, config.max_2d_position_embeddings-1]`. Each bounding box should be a normalized version in (x0, y0, x1, y1) format, where (x0, y0) corresponds to the position of the upper left corner in the bounding box, and (x1, y1) represents the position of the lower right corner. See [Overview](#Overview) for normalization. Examples: ```python >>> from transformers import AutoTokenizer, AutoModel >>> from datasets import load_dataset >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> model = AutoModel.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True) >>> example = dataset[0] >>> words = example["tokens"] >>> boxes = example["bboxes"] >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt") >>> outputs = model(**encoding) >>> last_hidden_states = outputs.last_hidden_state ```NzDYou cannot specify both input_ids and inputs_embeds at the same timer"z5You have to specify either input_ids or inputs_embeds)rC)rErM)rLr!rMrN)r}r!)rrrr(r)rr)r1 pooler_outputrr2)r=rr(use_return_dictr%warn_if_padding_and_no_attention_maskrIrEr7rJrKonesrrPrMr9get_extended_attention_mask get_head_maskr#rarbrcr rr2)r<rLr}rrMr!rrNrr(r)rO batch_sizerrEbuffered_token_type_ids buffered_token_type_ids_expandedextended_attention_maskembedding_outputlayout_embedding_outputencoder_outputssequence_outputrHs r?rQLiltModel.forwardrseP2C1N-TXT_T_TqTq$8$D $++JjJj &1%cd d  "  6 6y Q#..*K  &',,.s3KTU U!, %.%:!! @T@T <;;{T1FSD  !"ZZ*)A6RN  !t(899*.//*H*HKZK*X'3J3Q3QR\3i0!A!&[ SY!Z150P0PQ_0m&&y++2O2OP )-%)' *9* &,#'"8"8d"8"^,,  #2/!5#' *!,8< 8O O4UY #]3oab6II I)-')77&11   rA)r=rPrbrarc)T) NNNNNNNNNN)rarbrcrdr'rirlrqrrr7rrrrr rQrerfrgs@r?r_r_Qs$"/0C-1'+1515/3,004,0/3&*p ELL)p u||$p !. p !. p u||, p ELL)p   -p $D>p 'tnp d^p  uU\\"$>> ?p p rAr_z LiLT Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled output) e.g. for GLUE tasks. ) custom_introc^\rSrSrU4Sjr\SS\\RS\\RS\\RS\\RS\\RS\\RS \\RS \\RS \\ S \\ S \\ S\ \ \R\44Sjj5rSrU=r$)LiltForSequenceClassificationic>[TU]U5 URUlXl[ USS9Ul[ U5UlUR5 gNF)re) r&r' num_labelsr=r_rLLiltClassificationHead classifierrdr;s r?r'&LiltForSequenceClassification.__init__sH   ++ f> 08 rArLr}rrMr!rrNlabelsrr(r)rc hU bU OURRn URUUUUUUUU U U S9 n U Sn URU 5nSnUGbUR UR 5nURR cURS:XaSURlOoURS:aNUR[R:XdUR[R:XaSURlOSURlURR S:XaI[5nURS:Xa&U"UR5UR55nOU"X5nOURR S:Xa=[5nU"URSUR5URS55nO,URR S:Xa[!5nU"X5nU (dU4U S S-nUbU4U-$U$[#UUU R$U R&S 9$) a$ bbox (`torch.LongTensor` of shape `(batch_size, sequence_length, 4)`, *optional*): Bounding boxes of each input sequence tokens. Selected in the range `[0, config.max_2d_position_embeddings-1]`. Each bounding box should be a normalized version in (x0, y0, x1, y1) format, where (x0, y0) corresponds to the position of the upper left corner in the bounding box, and (x1, y1) represents the position of the lower right corner. See [Overview](#Overview) for normalization. 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 AutoTokenizer, AutoModelForSequenceClassification >>> from datasets import load_dataset >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> model = AutoModelForSequenceClassification.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True) >>> example = dataset[0] >>> words = example["tokens"] >>> boxes = example["bboxes"] >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt") >>> outputs = model(**encoding) >>> predicted_class_idx = outputs.logits.argmax(-1).item() >>> predicted_class = model.config.id2label[predicted_class_idx] ```N r}rrMr!rrNrr(r)rr regressionsingle_label_classificationmulti_label_classificationr"rzlosslogitsrr2)r=rvrLrrGrE problem_typerrDr7rKrWr squeezerrrrrr2r<rLr}rrMr!rrNrrr(r)rrrrloss_fctrs r?rQ%LiltForSequenceClassification.forwards\&1%` 'tn` d^`  uU\\"$<< =` ` rArc^\rSrSrU4Sjr\SS\\RS\\RS\\RS\\RS\\RS\\RS \\RS \\RS \\ S \\ S \\ S\ \ \R\44Sjj5rSrU=r$)LiltForTokenClassificationi]cd>[TU]U5 URUl[USS9UlUR b UR O UR n[R"U5Ul [R"URUR5Ul UR5 gr)r&r'rr_rLclassifier_dropoutr4rr3r5rvr*rrdr<r=rr>s r?r'#LiltForTokenClassification.__init__`s   ++f> )/)B)B)NF % %TZTnTn zz"45 ))F$6$68I8IJ rArLr}rrMr!rrNrrr(r)rc U bU OURRn URUUUUUUUU U U S9 n U Sn URU 5n UR U 5nSnUbWUR UR 5n[5nU"URSUR5URS55nU (dU4U SS-nUbU4U-$U$[UUU RU RS9$)a8 bbox (`torch.LongTensor` of shape `(batch_size, sequence_length, 4)`, *optional*): Bounding boxes of each input sequence tokens. Selected in the range `[0, config.max_2d_position_embeddings-1]`. Each bounding box should be a normalized version in (x0, y0, x1, y1) format, where (x0, y0) corresponds to the position of the upper left corner in the bounding box, and (x1, y1) represents the position of the lower right corner. See [Overview](#Overview) for normalization. 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 AutoTokenizer, AutoModelForTokenClassification >>> from datasets import load_dataset >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> model = AutoModelForTokenClassification.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True) >>> example = dataset[0] >>> words = example["tokens"] >>> boxes = example["bboxes"] >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt") >>> outputs = model(**encoding) >>> predicted_class_indices = outputs.logits.argmax(-1) ```Nrrr"rzr) r=rvrLr5rrGrErrrrrr2rs r?rQ"LiltForTokenClassification.forwardns V&1%N 'tnN d^N  uU\\"$99 :N N rArc2^\rSrSrSrU4SjrSrSrU=r$)riz-Head for sentence-level classification tasks.cb>[TU]5 [R"URUR5UlUR b UR O URn[R"U5Ul [R"URUR5Ul gr) r&r'rrvr*rrr4r3r5rout_projrs r?r'LiltClassificationHead.__init__s YYv1163E3EF )/)B)B)NF % %TZTnTn zz"45  &"4"4f6G6GH rAc USS2SSS24nURU5nURU5n[R"U5nURU5nUR U5nU$rF)r5rr7tanhr)r<featureskwargsrs r?rQLiltClassificationHead.forwardsY Q1W  LLO JJqM JJqM LLO MM! rA)rr5r) rarbrcrd__doc__r'rQrerfrgs@r?rrs7IrArc^\rSrSrU4Sjr\SS\\RS\\RS\\RS\\RS\\RS\\RS \\RS \\RS \\RS \\ S \\ S\\ S\ \ \R\44Sjj5rSrU=r$)LiltForQuestionAnsweringic>[TU]U5 URUl[USS9Ul[ R "URUR5UlUR5 gr) r&r'rr_rLrrvr* qa_outputsrdr;s r?r'!LiltForQuestionAnswering.__init__sU   ++f> ))F$6$68I8IJ rArLr}rrMr!rrNstart_positions end_positionsrr(r)rc (U bU OURRn URUUUUUUUU U U S9 n U SnURU5nUR SSS9unnUR S5R 5nUR S5R 5nSnUbU b[UR55S:aUR S5n[U R55S:aU R S5n URS5nURSU5nU RSU5n [US9nU"UU5nU"UU 5nUU-S- nU (dUU4U SS-nUbU4U-$U$[UUUU RU RS 9$) aB bbox (`torch.LongTensor` of shape `(batch_size, sequence_length, 4)`, *optional*): Bounding boxes of each input sequence tokens. Selected in the range `[0, config.max_2d_position_embeddings-1]`. Each bounding box should be a normalized version in (x0, y0, x1, y1) format, where (x0, y0) corresponds to the position of the upper left corner in the bounding box, and (x1, y1) represents the position of the lower right corner. See [Overview](#Overview) for normalization. Examples: ```python >>> from transformers import AutoTokenizer, AutoModelForQuestionAnswering >>> from datasets import load_dataset >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> model = AutoModelForQuestionAnswering.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base") >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True) >>> example = dataset[0] >>> words = example["tokens"] >>> boxes = example["bboxes"] >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt") >>> 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] >>> predicted_answer = tokenizer.decode(predict_answer_tokens) ```Nrrrr"rT) ignore_indexrz)r start_logits end_logitsrr2)r=rvrLrsplitrrrrIclamprrrr2)r<rLr}rrMr!rrNrrrr(r)rrrrr total_loss ignored_indexr start_lossend_lossrs r?rQ LiltForQuestionAnswering.forwards^&1%_ 'tn_ d^_  uU\\"$@@ A_ _ rAr)rrrr_rK)5rrtypingrrrr7torch.utils.checkpointrtorch.nnrrr activationsr modeling_outputsr r rrrmodeling_utilsr pytorch_utilsrrrutilsrrconfiguration_liltr get_loggerraloggerModulerrirrrrrrrr@rKr_rrrr__all__r+rAr?rs )) AA!.ll,*   H %V=V=r5+2995+pF FT RYY 1BII1j ryy   9 9xD "))D P  */**.Q #Q Q h n $7n  n b_ !4_ _ FRYY,l 2l l ^ rA