`iDdZddlZddlmZddlmZmZddlZddlmZddl m Z ddl m Z dd l mZdd lmZmZmZdd lmZdd lmZmZmZdd lmZmZmZmZddlmZej e!Z"eedGddeZ#Gddej$Z%Gddej$Z&Gddej$Z'Gddej$Z(Gddej$Z)Gddej$Z*Gd d!ej$Z+Gd"d#ej$Z,Gd$d%ej$Z-Gd&d'eZ.Gd(d)ej$Z/Gd*d+ej$Z0Gd,d-ej$Z1eGd.d/eZ2eGd0d1e2Z3eGd2d3e2Z4ed4Gd5d6e2Z5ed7Gd8d9e2Z6gd:Z7dS);zPyTorch Bros model.N) dataclass)OptionalUnion)nn)CrossEntropyLoss)ACT2FN)GradientCheckpointingLayer)"BaseModelOutputWithCrossAttentions,BaseModelOutputWithPoolingAndCrossAttentionsTokenClassifierOutput)PreTrainedModel)apply_chunking_to_forward find_pruneable_heads_and_indicesprune_linear_layer) ModelOutputauto_docstringcan_return_tuplelogging) BrosConfigz@ Base class for outputs of token classification models. ) custom_introceZdZUdZdZeejed<dZ eejed<dZ eejed<dZ ee ejed<dZ ee ejed<dS)BrosSpadeOutputa loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Classification loss. initial_token_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.num_labels)`): Classification scores for entity initial tokens (before SoftMax). subsequent_token_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, sequence_length+1)`): Classification scores for entity sequence tokens (before SoftMax). Nlossinitial_token_logitssubsequent_token_logits hidden_states attentions)__name__ __module__ __qualname____doc__rrtorch FloatTensor__annotations__rrrtuplerz/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/transformers/models/bros/modeling_bros.pyrr)s)-D(5$ %,,,8<(5#45<<<;?Xe&78???8> B||BB22T]5G5G1aQUQnrsQs5t5tt )\--//1A1A1C1CD"MMMr)r r!r"r1r$TensorrJ __classcell__r7s@r*r,r,@s^33333u| r)r,cBeZdZfdZdejdejfdZxZS)BrosPositionalEmbedding2Dct|j|_t||_t||_dSN)r0r1dim_bboxr, x_pos_emb y_pos_embr5r6r7s r*r1z"BrosPositionalEmbedding2D.__init__VsD  26::26::r)bboxr9c8g}t|jD]l}|dzdkr1|||d|f<|||d|fmt j|d}|S)Nr;r.r<r=)rangerSappendrTrUr$rA)r5rWstacki bbox_pos_embs r*rJz!BrosPositionalEmbedding2D.forward]st}%% ; ;A1uzz T^^DaL99:::: T^^DaL99::::yB/// r)rKrNs@r*rPrPUs^;;;;;ELU\r)rPc4eZdZfdZdejfdZxZS)BrosBboxEmbeddingsctt||_t j|j|jd|_dS)NF)bias) r0r1rPbbox_sinusoid_embrLineardim_bbox_sinusoid_emb_2ddim_bbox_projectionbbox_projectionrVs r*r1zBrosBboxEmbeddings.__init__isO !:6!B!B!y)H&Jdkpqqqr)rWc|dd}|dddddddf|dddddddfz }||}||}|S)Nrr) transposerbrf)r5rWbbox_tbbox_posr]s r*rJzBrosBboxEmbeddings.forwardns}1%%$111aaa-(6!!!T111aaa-+@@--h77 ++L99 r)rKrNs@r*r_r_hsZrrrrr ELr)r_c eZdZdZfdZ d deejdeejdeejdeejdejf d ZxZ S) BrosTextEmbeddingszGConstruct the embeddings from word, position and token_type embeddings.cttj|j|j|j|_tj|j|j|_ tj|j |j|_ tj |j|j |_ tj|j|_t#|dd|_|dt)j|jd|dt)j|jt(j|jjd dS) N) padding_idxepsposition_embedding_typeabsolute position_ids)rr<token_type_idsdtypedeviceF) persistent)r0r1r Embedding vocab_size hidden_size pad_token_idword_embeddingsmax_position_embeddingsposition_embeddingstype_vocab_sizetoken_type_embeddings LayerNormlayer_norm_epsDropouthidden_dropout_probdropoutgetattrrqr4r$r3expandzerosrsr?longrwrVs r*r1zBrosTextEmbeddings.__init__zsN !|F,=v?Q_e_rsss#%<0NPVPb#c#c %'\&2H&J\%]%]"f&8f>STTTz&"<== '.v7PR\']']$ ^U\&:X-Y-Y-`-`ah-i-ijjj   K!&&((j(/          r)N input_idsrtrs inputs_embedsr9c||}n|dd}|d}||jddd|f}|mt|dr2|jddd|f}||d|}|}n+t j|t j|jj}|| |}| |} || z} |j dkr| |} | | z } | | } || } | S)Nr<rrtrrurr)r?rshasattrrtrr$rrrwr}rrqrrr) r5rrtrsr input_shape seq_lengthbuffered_token_type_ids buffered_token_type_ids_expandedr embeddingsrs r*rJzBrosTextEmbeddings.forwardsb  #..**KK',,..ss3K ^  ,QQQ  ^ J J"%::  ': 5 5"&":":<"H"H  - -J^^J// \\*-- r))NNNN) r r!r"r#r1rr$rLrJrMrNs@r*rlrlwsQQ     4-115/304 ##EL)#!.#u|, #   - #  ########r)rlceZdZfdZ d dejdejdeejdeejdeejd eejd eejd eejfd ZxZ S)BrosSelfAttentionc@t|j|jzdkr0t |ds t d|jd|jd|j|_t |j|jz |_|j|jz|_tj |j|j|_ tj |j|j|_ tj |j|j|_ tj|j|_t#|dd|_|jdks |jd kr6|j|_tjd |jzd z |j|_|j|_dS) Nrembedding_sizezThe hidden size (z6) is not a multiple of the number of attention heads ()rqrr relative_keyrelative_key_queryr;r)r0r1r{num_attention_headsr ValueErrorintattention_head_size all_head_sizerrcquerykeyvaluerattention_probs_dropout_probrrrqr~rydistance_embedding is_decoderrVs r*r1zBrosSelfAttention.__init__s    : :a ? ?PVXhHiHi ?8F$688 4888  $*#= #&v'9F 9 9T=Y]q=q=q+1+ID (&(l1v7U3UXY3Y[_[s&t&tD # +r)NFrr]attention_mask head_maskencoder_hidden_statesencoder_attention_maskoutput_attentionsr9c|jdd|j|jf}|||dd} |du} | r{|||dd} |||dd} |}nx|||dd} |||dd} tj | | dd} |j dks |j dkr4| d}tj |tj |jdd}tj |tj |jdd}||z }|||jzdz }|| j }|j dkrtjd | |}| |z} n?|j dkr4tjd | |}tjd | |}| |z|z} | j\}}}}|||||}|gd }tjd | |f}| |z} | t+j|jz } || |z} t/jd| }||}|||z}tj || }|dddd}| dd|jfz}|j|}|r||fn|f}|jr|dz}|S)Nrr<rr;rrru)rvzbhld,lrd->bhlrzbhrd,lrd->bhlr)r;rrrzbnid,bijd->bnijr=rrR)shaperrrr@rhrrr$matmulrqr?r3rrwrr~torveinsumpermutemathsqrtrSoftmaxr contiguousrr)r5rr]rrrrr hidden_shape query_layeris_cross_attention key_layer value_layerattention_scoresrposition_ids_lposition_ids_rdistancepositional_embeddingrelative_position_scoresrelative_position_scores_queryrelative_position_scores_key batch_sizen_headd_headbbox_pos_scoresattention_probs context_layernew_context_layer_shapeoutputss r*rJzBrosSelfAttention.forwards&+A.D4LdNfg jj//44\BBLLQPQRR 3$>  W!677<<\JJTTUVXYZZI**%:;;@@NNXXYZ\]^^K3NN//44\BBLLQPQRRI**]3388FFPPQRTUVVK!< Y5H5HR5P5PQQ  '> 9 9T=Y]q=q=q&++--a0J"\*EJ}OcdddiijlnoppN"\*EJ}OcdddiijkmoppN%6H#'#:#:8dFb;bef;f#g#g #7#:#:AR#:#S#S +~==+0<8H+Wk+l+l(#36N#N  -1EEE16>NP[]q1r1r./4|>,,77  - 9O _kBB %--aAq99DDFF "/"4"4"6"6ss";t?Q>S"S* *,CD 6G]=/22mM] ? ('Gr)NNNNF) r r!r"r1r$rLrr'rJrMrNs@r*rrs,,,,,826,08<9=49PP|PlP!. P EL) P ( 5 P!) 6P$EL1P u| PPPPPPPPr)rcPeZdZfdZdejdejdejfdZxZS)BrosSelfOutputcttj|j|j|_tj|j|j|_tj|j |_ dSNro) r0r1rrcr{denserrrrrrVs r*r1zBrosSelfOutput.__init__&sf Yv163EFF f&8f>STTTz&"<== r)r input_tensorr9c||}||}|||z}|SrRrrrr5rrs r*rJzBrosSelfOutput.forward,@ =11  ]33 }|'CDD r)rKrNs@r*rr%i>>>>> U\RWR^r)rceZdZfdZdZ ddejdejdeejdeejd eejd eejd eed e ejfd Z xZ S) BrosAttentionctt||_t ||_t |_dSrR)r0r1rr5routputset pruned_headsrVs r*r1zBrosAttention.__init__4sI %f-- $V,, EEr)ct|dkrdSt||jj|jj|j\}}t |jj||j_t |jj||j_t |jj ||j_ t |j j |d|j _ |jjt|z |j_|jj|jjz|j_ |j ||_dS)Nrrr=)lenrr5rrrrrrrrrrunion)r5headsindexs r* prune_headszBrosAttention.prune_heads:s u::?? F7  I ) I )     u-TY_eDD *49=%@@ ,TY_eDD .t{/@%QOOO )- (EE (R %"&)"?$)B_"_  -33E::r)NFrr]rrrrrr9c ||||||||}||d|} | f|ddz} | S)Nrr]rrrrrrr)r5r) r5rr]rrrrr self_outputsattention_outputrs r*rJzBrosAttention.forwardOsfyy'%)"7#9/!   ;;|A FF#% QRR(88r)r) r r!r"r1rr$rLrboolr'rJrMrNs@r*rr3s""""" ;;;226,08<9=,1|l!.  EL)  ( 5 !) 6$D> u| r)rcBeZdZfdZdejdejfdZxZS)BrosIntermediatecttj|j|j|_t|jtrt|j|_ dS|j|_ dSrR) r0r1rrcr{intermediate_sizer isinstance hidden_actstrr intermediate_act_fnrVs r*r1zBrosIntermediate.__init__isn Yv163KLL f' - - 9'-f.?'@D $ $ $'-'8D $ $ $r)rr9cZ||}||}|SrR)rr)r5rs r*rJzBrosIntermediate.forwardqs, =11 00?? r)rKrNs@r*rrhs^99999U\elr)rcPeZdZfdZdejdejdejfdZxZS) BrosOutputcttj|j|j|_tj|j|j|_tj |j |_ dSr) r0r1rrcrr{rrrrrrrVs r*r1zBrosOutput.__init__xsf Yv79KLL f&8f>STTTz&"<== r)rrr9c||}||}|||z}|SrRrrs r*rJzBrosOutput.forward~rr)rKrNs@r*rrwrr)rceZdZfdZ ddejdejdeejdeejdeejd eejd eed e ejfd Z d Z xZ S) BrosLayerc~t|j|_d|_t ||_|j|_|j|_|jr-|jst|dt ||_ t||_ t||_ dS)Nrz> should be used as a decoder model if cross attention is added)r0r1chunk_size_feed_forward seq_len_dimr attentionradd_cross_attention Exceptioncrossattentionr intermediaterrrVs r*r1zBrosLayer.__init__s '-'E$&v.. +#)#=  # 8? i4 g g ghhh"/"7"7D ,V44 (( r)NFrr]rrrrrr9c||||||}|d} |jr |dd} n |dd} |jrU|St|drtd|d|| |||||} | d} | | ddz} t |j|j|j| } | f| z} |jr| d z} | S) N)r]rrrrrr<rz'If `encoder_hidden_states` are passed, z` has to be instantiated with cross-attention layers by setting `config.add_cross_attention=True`)rrrrrrR) rrrrrrfeed_forward_chunkrr) r5rr]rrrrrself_attention_outputsrrcross_attention_outputs layer_outputs r*rJzBrosLayer.forwardsO"& %)/ "0" " 2!4 ? 1,QrT2GG,QRR0G ? >4@t-.. edeee'+&9&9 -#&;'="3 ':'' # 7q9  7" ==G0  #  (       /G+ ? ('Gr)c\||}|||}|SrR)rr)r5rintermediate_outputrs r*rzBrosLayer.feed_forward_chunks2"//0@AA{{#68HII r)r) r r!r"r1r$rLrr%rr'rJrrMrNs@r*rrs ) ) ) ) )$7;15=A>B,166|6l6!!23 6 E-. 6 ((9: 6!)): ;6$D>6 u| 6666pr)rc$eZdZfdZe ddejdejdeejdeejd eejd eejd ee d ee d ee de e eje ffdZ xZS) BrosEncoderct|_tjfdt jD|_dS)Nc.g|]}tSr()r).0_r6s r* z(BrosEncoder.__init__..s!#_#_#_!If$5$5#_#_#_r))r0r1r6r ModuleListrYnum_hidden_layerslayerrVs `r*r1zBrosEncoder.__init__sV  ]#_#_#_#_uVE]?^?^#_#_#_`` r)NFTrr]rrrrroutput_hidden_states return_dictr9c P|rdnd} |rdnd} |r|jjrdnd} t|jD]Y\} }|r| |fz} ||| nd}||||||||}|d}|r$| |dfz} |jjr | |dfz} Z|r| |fz} t || | | S)Nr(rrrr;)last_hidden_staterrcross_attentions)r6r enumeraterr )r5rr]rrrrrrrall_hidden_statesall_self_attentionsall_cross_attentionsr\ layer_modulelayer_head_mask layer_outputss r*rJzBrosEncoder.forwards4#7@BBD$5?bb4%6d4;;Zdrr`d(44 V VOA|# I$58H$H!.7.CillO(L+)-)&;'="3M*!,M  V&9]1=MB,1/4&*. . |. l. !!23 . E-. . ((9: . !)): ;. $D>. 'tn. d^.  uU\"$FF G. . . . . . . . r)rcBeZdZfdZdejdejfdZxZS) BrosPoolercttj|j|j|_tj|_dSrR)r0r1rrcr{rTanh activationrVs r*r1zBrosPooler.__init__ sC Yv163EFF '))r)rr9cr|dddf}||}||}|S)Nr)rr )r5rfirst_token_tensor pooled_outputs r*rJzBrosPooler.forwards@+111a40 #566  66 r)rKrNs@r*rr s^$$$$$ U\elr)rcBeZdZfdZdejdejfdZxZS)BrosRelationExtractorct|j|_|j|_|j|_|j|_tj|j|_ tj |j|j|jz|_ tj |j|j|jz|_ tj tjd|j|_dS)Nr)r0r1 n_relationsr{backbone_hidden_sizehead_hidden_sizeclassifier_dropout_probrrdroprcrr Parameterr$r dummy_noderVs r*r1zBrosRelationExtractor.__init__s !-$*$6! & 2'-'E$Jt;<< Yt8$:JTMb:bcc 9T68H4K`8`aa,u{1d6O'P'PQQr)rrc |||}|jdd|dd}t j||gd}|||}| |d|d|j |j }| |d|d|j |j }t j | dddd| dddd}|S)Nrraxisr;r)rr+r- unsqueezerepeatr?r$rArr@r'r)rr)r5rr dummy_vecrelation_scores r*rJzBrosRelationExtractor.forward)sFjj;!7!788 O--a00779>>!;L;LaPP Iy)41=== HHTYYy1122 !&&   Q  !1!1!!4!4d6FH]  NN9>>!#4#4innQ6G6GIY[_[pqq    1a + +Y->->q!Q-J-J  r)rKrNs@r*r%r%sc R R R R R5<ELr)r%c4eZdZUeed<dZdejfdZdS)BrosPreTrainedModelr6brosmodulec|jj}t|tjrJ|jjd||j |jj dSdSt|tj rU|jjd||j +|jj|j  dSdSt|tj r?|jj |jj ddSt|tr(tj|j|dSdS)zInitialize the weightsr.)meanstdNg?)r;)r6initializer_rangerrrcweightdatanormal_razero_ryrnrfill_r%initr-)r5r8r;s r* _init_weightsz!BrosPreTrainedModel._init_weightsAs]k+ fbi ( ( 8 M  & &CS & 9 9 9{&  &&((((('&  - - 8 M  & &CS & 9 9 9!- "6#56<<>>>>>.-  - - 8 K  " " $ $ $ M  $ $S ) ) ) ) )  5 6 6 8 GOOF-3O 7 7 7 7 7 8 8r)N) r r!r"rr&base_model_prefixrModulerCr(r)r*r6r6<sE 8BI888888r)r6ceZdZdfd ZdZdZdZee dde e j de e j d e e j d e e j d e e j d e e j d e e j de e j de e j de e de e de e de ee j effdZxZS) BrosModelTc(t|||_t||_t ||_t||_|rt|nd|_ | dS)zv add_pooling_layer (bool, *optional*, defaults to `True`): Whether to add a pooling layer N) r0r1r6rlrr_bbox_embeddingsrencoderrpooler init_weights)r5r6add_pooling_layerr7s r*r1zBrosModel.__init__Ws     ,V441&99"6** ,=Gj(((4  r)c|jjSrRrr})r5s r*get_input_embeddingszBrosModel.get_input_embeddingsgs ..r)c||j_dSrRrO)r5rs r*set_input_embeddingszBrosModel.set_input_embeddingsjs*/'''r)c|D]/\}}|jj|j|0dS)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)itemsrJrrr)r5heads_to_prunerrs r* _prune_headszBrosModel._prune_headsmsU +0022 C CLE5 L u % / ; ;E B B B B C Cr)NrrWrrtrsrrrrrrrr9c | | n |jj} | | n |jj} | | n |jj} ||t d||} n.||dd} nt d|t d| \}}||jn|j}|tj| |}|gt|j dr1|j j ddd|f}| ||}|}n!tj | tj|}||| |}|jjrL|J|\}}}||f}| tj||} || }nd}|||jj}| |||| }|jdd kr|ddddgd f}||jjz}||}|||||||| | d }|d}|j||nd}t3|||j|j|jS)a bbox ('torch.FloatTensor' of shape '(batch_size, num_boxes, 4)'): Bounding box coordinates for each token in the input sequence. Each bounding box is a list of four values (x1, y1, x2, y2), where (x1, y1) is the top left corner, and (x2, y2) is the bottom right corner of the bounding box. Examples: ```python >>> import torch >>> from transformers import BrosProcessor, BrosModel >>> processor = BrosProcessor.from_pretrained("jinho8345/bros-base-uncased") >>> model = BrosModel.from_pretrained("jinho8345/bros-base-uncased") >>> encoding = processor("Hello, my dog is cute", add_special_tokens=False, return_tensors="pt") >>> bbox = torch.tensor([[[0, 0, 1, 1]]]).repeat(1, encoding["input_ids"].shape[-1], 1) >>> encoding["bbox"] = bbox >>> 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_embedszYou have to specify bbox)rwrtru)rrsrtr)rrr;rr;rrrT)r]rrrrrrrr)r pooler_outputrrr)r6rruse_return_dictrr?rwr$onesrrrtrrrget_extended_attention_maskrinvert_attention_mask get_head_maskrr bbox_scalerIrJrKr rrr)r5rrWrrtrsrrrrrrrrrrrwrrextended_attention_maskencoder_batch_sizeencoder_sequence_lengthr encoder_hidden_shapeencoder_extended_attention_maskembedding_output scaled_bboxbbox_position_embeddingsencoder_outputssequence_outputr#s r*rJzBrosModel.forwardus P2C1N--TXT_Tq$8$D $+Jj &1%cdd d  "#..**KK  &',,..ss3KKTUU U <788 8!, J%.%:!! @T  !"Z FCCCN  !t(899 [*./*HKZK*X'3J3Q3QR\^h3i3i0!A!&[ SY!Z!Z!Z150P0PQ_alnt0u0u ; ! 3&;&G=R=W=W=Y=Y :  7$68O#P %-).4HQW)X)X)X&.2.H.HI_.`.` + +.2 +&&y$+2OPP ??%)' +   :b>Q  11166667DT[33 #'#7#7 #D#D ,, 12"7#B/!5'   *!,8< 8O O444UY ;-')7&1,=     r))T NNNNNNNNNNNN)r r!r"r1rPrRrVrrrr$rLrrr'r rJrMrNs@r*rGrGUs ///000CCC-1'+1515/3,0048<9=,0/3&*} } EL)} u|$} !. } !. } u|, } EL)}   -}  ( 5} !) 6} $D>} 'tn} d^}  uU\"$PP Q} } } ^} } } } } r)rGceZdZdgZfdZee ddeej deej deej deej deej d eej d eej d eej d eej d ee dee dee de e ej e ffdZxZS)BrosForTokenClassificationrKcht||j|_t||_t |dr|jn|j}tj ||_ tj |j |j|_ |dSNclassifier_dropout)r0r1 num_labelsrGr7rrorrrrrcr{ classifierrLr5r6ror7s r*r1z#BrosForTokenClassification.__init__s     +f%% )09M)N)N nF % %TZTn z"455 )F$68IJJ r)NrrWrbbox_first_token_maskrtrsrrlabelsrrrr9c J| | n |jj} ||||||||| | d } | d}||}||}d}| t }|Z|d}||d|j|| d|}n8||d|j| d}t||| j | j S)a bbox ('torch.FloatTensor' of shape '(batch_size, num_boxes, 4)'): Bounding box coordinates for each token in the input sequence. Each bounding box is a list of four values (x1, y1, x2, y2), where (x1, y1) is the top left corner, and (x2, y2) is the bottom right corner of the bounding box. bbox_first_token_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to indicate the first token of each bounding box. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. Examples: ```python >>> import torch >>> from transformers import BrosProcessor, BrosForTokenClassification >>> processor = BrosProcessor.from_pretrained("jinho8345/bros-base-uncased") >>> model = BrosForTokenClassification.from_pretrained("jinho8345/bros-base-uncased") >>> encoding = processor("Hello, my dog is cute", add_special_tokens=False, return_tensors="pt") >>> bbox = torch.tensor([[[0, 0, 1, 1]]]).repeat(1, encoding["input_ids"].shape[-1], 1) >>> encoding["bbox"] = bbox >>> outputs = model(**encoding) ```NT) rWrrtrsrrrrrrr<rlogitsrr) r6rZr7rrqrr@rpr rr)r5rrWrrsrtrsrrrtrrrrrirwrloss_fcts r*rJz"BrosForTokenClassification.forwards@Z&1%O 'tnO d^O  uU\"$99 :O O O ^O O O O O r)rla Bros Model with a token classification head on top (initial_token_layers and subsequent_token_layer on top of the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks. The initial_token_classifier is used to predict the first token of each entity, and the subsequent_token_classifier is used to predict the subsequent tokens within an entity. Compared to BrosForTokenClassification, this model is more robust to serialization errors since it predicts next token from one token. c!eZdZdgZfdZee ddeej deej deej deej deej d eej d eej d eej d eej d eej dee dee dee de e ej e ffdZxZS)!BrosSpadeEEForTokenClassificationrKc Ht|||_|j|_|j|_|j|_t||_t|dr|j n|j }tj tj|tj|j|jtj|tj|j|j|_t#||_|dSrn)r0r1r6rpr'r{r(rGr7rrorr Sequentialrrcinitial_token_classifierr%subsequent_token_classifierrLrrs r*r1z*BrosSpadeEEForTokenClassification.__init__hs      +!-$*$6!f%% )09M)N)N nF % %TZTn  )+ J) * * If(&*< = = J) * * If(&*; < < ) ) %,A+H+H( r)NrrWrrsrtrsrrinitial_token_labelssubsequent_token_labelsrrrr9c  | | n |jj} ||||||||| | d }|d}|dd}||dd}|||d}d|z }|j\}}|j }tj |tj |dg |gd}||dddddftj|jj}tj||dz |tj}||dddddftj|jj}|d}d}| | t+}| d} |G|d}||d|j|| |}n%||d|j| }| d} ||d|dz|| |}||z}t/||||j|j S) a> bbox ('torch.FloatTensor' of shape '(batch_size, num_boxes, 4)'): Bounding box coordinates for each token in the input sequence. Each bounding box is a list of four values (x1, y1, x2, y2), where (x1, y1) is the top left corner, and (x2, y2) is the bottom right corner of the bounding box. bbox_first_token_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to indicate the first token of each bounding box. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. initial_token_labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for the initial token classification. subsequent_token_labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for the subsequent token classification. Examples: ```python >>> import torch >>> from transformers import BrosProcessor, BrosSpadeEEForTokenClassification >>> processor = BrosProcessor.from_pretrained("jinho8345/bros-base-uncased") >>> model = BrosSpadeEEForTokenClassification.from_pretrained("jinho8345/bros-base-uncased") >>> encoding = processor("Hello, my dog is cute", add_special_tokens=False, return_tensors="pt") >>> bbox = torch.tensor([[[0, 0, 1, 1]]]).repeat(1, encoding["input_ids"].shape[-1], 1) >>> encoding["bbox"] = bbox >>> outputs = model(**encoding) ```NT rrWrrtrsrrrrrrrr/rwrvr<)rrrrr)r6rZr7rhrrrsqueezerrwr$rArrr masked_fillfinforvmineyer@rrprrr)r5rrWrrsrtrsrrrrrrrrlast_hidden_statesrrinv_attention_maskrmax_seq_lengthrwinvalid_token_maskself_token_masksubsequent_token_maskrrxinitial_token_losssubsequent_token_losss r*rJz)BrosSpadeEEForTokenClassification.forwardsAd&1% >  +0G0S'))H$8#<#?&&"" &.X.B.G.GDO.\.\^r%s%s"&=&B&B2&F&F #$,H',,R!1CDDEZ['(=>%% ! &(==D!5$;!/)     r)) NNNNNNNNNNNNN)r r!r"rzr1rrrr$rLrrr'rrJrMrNs@r*r|r|\s+4&2-1'+158<15/3,0047;:>,0/3&*o o EL)o u|$o !. o ( 5 o !. o u|,o EL)o   -o 'u|4o "*%,!7o $D>o 'tno d^o  uU\"O3 4o o o ^o o o o o r)r|z Bros Model with a token classification head on top (a entity_linker layer on top of the hidden-states output) e.g. for Entity-Linking. The entity_linker is used to predict intra-entity links (one entity to another entity). ceZdZdgZfdZee ddeej deej deej deej deej d eej d eej d eej d eej d ee dee dee de e ej e ffdZxZS)!BrosSpadeELForTokenClassificationrKcTt|||_|j|_|j|_|j|_t||_t|dr|j n|j t||_ |dSrn)r0r1r6rpr'r{r(rGr7rrorr% entity_linkerrLrVs r*r1z*BrosSpadeELForTokenClassification.__init__s      +!-$*$6!f%% &-f6J&K&K k " "QWQk26:: r)NrrWrrsrtrsrrrtrrrr9c  | | n |jj} ||||||||| | d } | d}|dd}|||d}d}| et}|j\}}|j }tj ||dz |tj }|d}tj|tj|dgtj |gd }||dddddftj|jj}||dddddftj|jj}||d|dz|| d|}t+||| j| j S) a bbox ('torch.FloatTensor' of shape '(batch_size, num_boxes, 4)'): Bounding box coordinates for each token in the input sequence. Each bounding box is a list of four values (x1, y1, x2, y2), where (x1, y1) is the top left corner, and (x2, y2) is the bottom right corner of the bounding box. bbox_first_token_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to indicate the first token of each bounding box. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. Examples: ```python >>> import torch >>> from transformers import BrosProcessor, BrosSpadeELForTokenClassification >>> processor = BrosProcessor.from_pretrained("jinho8345/bros-base-uncased") >>> model = BrosSpadeELForTokenClassification.from_pretrained("jinho8345/bros-base-uncased") >>> encoding = processor("Hello, my dog is cute", add_special_tokens=False, return_tensors="pt") >>> bbox = torch.tensor([[[0, 0, 1, 1]]]).repeat(1, encoding["input_ids"].shape[-1], 1) >>> encoding["bbox"] = bbox >>> outputs = model(**encoding) ```NTrrrrr<rur/rv)r6rZr7rhrrrrrrwr$rrrr@rArrrrvrr rr)r5rrWrrsrtrsrrrtrrrrrrwrrxrrrwrmasks r*rJz)BrosSpadeELForTokenClassification.forward s'X&1%??Ev{{SUW[G\]]D$!/)     r)rjryrNs@r*rrs+4&     -1'+158<15/3,004)-,0/3&*Y Y EL)Y u|$Y !. Y ( 5 Y !. Y u|,Y EL)Y   -Y &Y $D>Y 'tnY d^Y  uU\"$99 :Y Y Y ^Y Y Y Y Y r)r)r6rGrlr|r)8r#r dataclassesrtypingrrr$rtorch.nnr activationsr modeling_layersr modeling_outputsr r r modeling_utilsr pytorch_utilsrrrutilsrrrrconfiguration_brosr get_loggerr loggerrrEr,rPr_rlrrrrrrrrr%r6rGrlr|r__all__r(r)r*rsf !!!!!!"""""""" %%%%%%!!!!!!999999 .-----llllllllllKKKKKKKKKKKK******  H % %  :::::k::   :" * &        >>>>>>>>Biiiii iiiZ     RY   11111BI111j     ry           JJJJJ*JJJZ5 5 5 5 5 ")5 5 5 r        BID88888/8880^ ^ ^ ^ ^ #^ ^ ^ Ba a a a a !4a a a HM M M M M (;M M M ` l l l l l (;l l  l ^   r)