o wik@sHdZddlZddlmZddlmZmZddlZddlZddlm Z ddl m Z ddl m Z dd lmZdd lmZmZmZdd lmZdd lmZmZmZdd lmZmZmZddlmZe e!Z"eeddGdddeZ#Gddde j$Z%Gddde j$Z&Gddde j$Z'Gddde j$Z(Gddde j$Z)Gddde j$Z*Gd d!d!e j$Z+Gd"d#d#e j$Z,Gd$d%d%e j$Z-Gd&d'd'eZ.Gd(d)d)e j$Z/Gd*d+d+e j$Z0Gd,d-d-e j$Z1eGd.d/d/eZ2eGd0d1d1e2Z3eGd2d3d3e2Z4ed4dGd5d6d6e2Z5ed7dGd8d9d9e2Z6gd:Z7dS);zPyTorch Bros model.N) dataclass)OptionalUnion)nn)CrossEntropyLoss)ACT2FN)GradientCheckpointingLayer))BaseModelOutputWithPastAndCrossAttentions,BaseModelOutputWithPoolingAndCrossAttentionsTokenClassifierOutput)PreTrainedModel)apply_chunking_to_forward find_pruneable_heads_and_indicesprune_linear_layer) ModelOutputauto_docstringlogging) BrosConfigz@ Base class for outputs of token classification models. ) custom_introc@steZdZUdZdZeejed<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__rrrtuplerr%r%c/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/transformers/models/bros/modeling_bros.pyr*s  rc2eZdZfddZdejdejfddZZS)BrosPositionalEmbedding1Dcs@t|j|_ddtd|jd|j}|d|dS)Nri'g@inv_freq)super__init__dim_bbox_sinusoid_emb_1dr!arangeregister_buffer)selfconfigr* __class__r%r&r,Ds z"BrosPositionalEmbedding1D.__init__pos_seqreturncCsX|}|\}}}||||d|jddd|jd}tj||gdd}|S)Nrdim)sizeviewr*r-r!catsincos)r0r4seq_sizeb1b2b3 sinusoid_inppos_embr%r%r&forwardNs  (z!BrosPositionalEmbedding1D.forwardrrrr,r!TensorrE __classcell__r%r%r2r&r(As  r(cr')BrosPositionalEmbedding2Dcs*t|j|_t||_t||_dSN)r+r,dim_bboxr( x_pos_emb y_pos_embr0r1r2r%r&r,Ws  z"BrosPositionalEmbedding2D.__init__bboxr5cCsdg}t|jD]!}|ddkr|||d|fq|||d|fqtj|dd}|S)Nr6r.r7r8)rangerKappendrLrMr!r<)r0rOstacki bbox_pos_embr%r%r&rE^s z!BrosPositionalEmbedding2D.forwardrFr%r%r2r&rIVs rIcs,eZdZfddZdejfddZZS)BrosBboxEmbeddingscs.tt||_tj|j|jdd|_dS)NF)bias) r+r,rIbbox_sinusoid_embrLineardim_bbox_sinusoid_emb_2ddim_bbox_projectionbbox_projectionrNr2r%r&r,js  zBrosBboxEmbeddings.__init__rOcCs\|dd}|dddddddf|dddddddf}||}||}|S)Nrr) transposerWr[)r0rObbox_tbbox_posrTr%r%r&rEos 8  zBrosBboxEmbeddings.forwardrFr%r%r2r&rUis rUc sheZdZdZfddZ     ddeejdeejdeejd eejd ed ejf d d Z Z S)BrosTextEmbeddingszGConstruct the embeddings from word, position and token_type embeddings.csttj|j|j|jd|_t|j|j|_ t|j |j|_ tj |j|j d|_ t|j|_t|dd|_|dt|jd|jdtj|jtj|jjdd d dS) N) padding_idxepsposition_embedding_typeabsolute position_ids)rr7token_type_idsdtypedeviceF) persistent)r+r,r Embedding vocab_size hidden_size pad_token_idword_embeddingsmax_position_embeddingsposition_embeddingstype_vocab_sizetoken_type_embeddings LayerNormlayer_norm_epsDropouthidden_dropout_probdropoutgetattrrcr/r!r.expandzerosrer:longrirNr2r%r&r,{s"  zBrosTextEmbeddings.__init__Nr input_idsrfre inputs_embedspast_key_values_lengthr5c Cs|dur |}n|dd}|d}|dur&|jdd|||f}|durPt|drE|jddd|f}||d|} | }n tj|tj|jjd}|durY| |}| |} || } |j dkrp| |} | | 7} | | } || } | S)Nr7rrfrrgrd)r:rehasattrrfrzr!r{r|rirorsrcrqrtrx) r0r}rfrer~r input_shape seq_lengthbuffered_token_type_ids buffered_token_type_ids_expandedrs embeddingsrqr%r%r&rEs,        zBrosTextEmbeddings.forward)NNNNr) rrrr r,rr!rGintrErHr%r%r2r&r_xs* r_cseZdZfddZdejfddZ      ddejd ejd eejd eejd eejd eejdeeeej deejdeejfddZ Z S)BrosSelfAttentioncst|j|jdkrt|dstd|jd|jd|j|_t|j|j|_|j|j|_t |j|j|_ t |j|j|_ t |j|j|_ t |j|_t|dd|_|jdksf|jd krw|j|_t d |jd |j|_|j|_dS) Nrembedding_sizezThe hidden size (z6) is not a multiple of the number of attention heads ()rcrd relative_keyrelative_key_queryr6r)r+r,rmnum_attention_headsr ValueErrorrattention_head_size all_head_sizerrXquerykeyvaluervattention_probs_dropout_probrxryrcrprkdistance_embedding is_decoderrNr2r%r&r,s&   zBrosSelfAttention.__init__xcCs6|dd|j|jf}|j|}|ddddS)Nr7rr6rr)r:rrr;permute)r0r new_x_shaper%r%r&transpose_for_scoress  z&BrosSelfAttention.transpose_for_scoresNFrrTattention_mask head_maskencoder_hidden_statesencoder_attention_maskpast_key_valueoutput_attentionsr5c Cs||} |du} | r|dur|d} |d} |}nP| r/|||} |||} |}n;|durZ|||} |||} tj|d| gdd} tj|d| gdd} n|||} |||} || } |jrv| | f}t| | dd}|j dks|j dkr| d}tj |tj |j d dd}tj |tj |j d dd}||}|||jd}|j| jd }|j dkrtd | |}||}n|j dkrtd | |}td | |}|||}| j\}}}}|||||}|gd }td| |f}||}|t|j}|dur||}tjdd|}||}|dur4||}t|| }|dddd}| dd|jf}|j|}|r\||fn|f}|jrh||f}|S)Nrrr6r8r7rrrg)rhzbhld,lrd->bhlrzbhrd,lrd->bhlr)r6rrrzbnid,bijd->bnijr)rrrrr!r<rmatmulr\rcr:r.r|rir;rrptorheinsumshapermathsqrtrrSoftmaxrx contiguousr)r0rrTrrrrrrmixed_query_layeris_cross_attention key_layer value_layer query_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_shapeoutputsr%r%r&rEsn              zBrosSelfAttention.forwardNNNNNF) rrrr,r!rGrrr$r"rErHr%r%r2r&rs8    rc8eZdZfddZdejdejdejfddZZS)BrosSelfOutputcsBtt|j|j|_tj|j|jd|_t|j |_ dSNra) r+r,rrXrmdensertrurvrwrxrNr2r%r&r,F zBrosSelfOutput.__init__r input_tensorr5cC&||}||}|||}|SrJrrxrtr0rrr%r%r&rEL  zBrosSelfOutput.forwardrFr%r%r2r&rE $rcseZdZfddZddZ      ddejdejd eejd eejd eejd eejd eeeej dee deejfddZ Z S) BrosAttentioncs*tt||_t||_t|_dSrJ)r+r,rr0routputset pruned_headsrNr2r%r&r,Ts    zBrosAttention.__init__cCst|dkrdSt||jj|jj|j\}}t|jj||j_t|jj||j_t|jj ||j_ t|j j |dd|j _ |jjt||j_|jj|jj|j_ |j ||_dS)Nrrr8)lenrr0rrrrrrrrrrunion)r0headsindexr%r%r& prune_headsZs zBrosAttention.prune_headsNFrrTrrrrrrr5c Cs@|j||||||||d} || d|} | f| dd} | S)N)rrTrrrrrrrr)r0r) r0rrTrrrrrr self_outputsattention_outputrr%r%r&rEos  zBrosAttention.forwardr) rrrr,rr!rGrr$r"boolrErHr%r%r2r&rSs8   rcr')BrosIntermediatecsDtt|j|j|_t|jt rt |j|_ dS|j|_ dSrJ) r+r,rrXrmintermediate_sizer isinstance hidden_actstrrintermediate_act_fnrNr2r%r&r,s   zBrosIntermediate.__init__rr5cCs||}||}|SrJ)rr)r0rr%r%r&rEs  zBrosIntermediate.forwardrFr%r%r2r&rs rcr) BrosOutputcsBtt|j|j|_tj|j|jd|_t |j |_ dSr) r+r,rrXrrmrrtrurvrwrxrNr2r%r&r,rzBrosOutput.__init__rrr5cCrrJrrr%r%r&rErzBrosOutput.forwardrFr%r%r2r&rrrcseZdZfddZ      ddejdejdeejdeejd eejd eejd eeeejd ee d eejfddZ ddZ Z S) BrosLayercsnt|j|_d|_t||_|j|_|j|_|jr+|js&t|dt||_ t ||_ t ||_ dS)Nrz> should be used as a decoder model if cross attention is added)r+r,chunk_size_feed_forward seq_len_dimr attentionradd_cross_attention Exceptioncrossattentionr intermediaterrrNr2r%r&r,s    zBrosLayer.__init__NFrrTrrrrrrr5c  Cs|dur |ddnd} |j|||||| d} | d} |jr)| dd} | d} n| dd} d}|jrp|durpt|drEtd|d|durO|d dnd}|| ||||||}|d} | |dd} |d}| |} t|j|j|j| }|f| } |jr| | f} | S) Nr6)rTrrrrrrr7rz'If `encoder_hidden_states` are passed, z` has to be instantiated with cross-attention layers by setting `config.add_cross_attention=True`r) rrrrrrfeed_forward_chunkrr)r0rrTrrrrrrself_attn_past_key_valueself_attention_outputsrrpresent_key_valuecross_attn_present_key_valuecross_attn_past_key_valuecross_attention_outputs layer_outputr%r%r&rEsX         zBrosLayer.forwardcCs||}|||}|SrJ)rr)r0rintermediate_outputrr%r%r&rs  zBrosLayer.feed_forward_chunkr) rrrr,r!rGrr"r$rrErrHr%r%r2r&rs8   ErcseZdZfddZ         ddejdejdeejd eejd eejd eejd eeeejd ee dee dee dee de eeje ffddZ Z S) BrosEncodercs4t|_tfddtjD|_dS)Ncsg|]}tqSr%)r).0_r1r%r& sz(BrosEncoder.__init__..)r+r,r1r ModuleListrPnum_hidden_layerslayerrNr2rr&r,s $zBrosEncoder.__init__NFTrrTrrrrpast_key_values use_cacheroutput_hidden_states return_dictr5c  Cs| rdnd} | r dnd} | r|jjrdnd}|rdnd}t|jD]L\}}| r,| |f} |dur4||nd}|dur>||nd}||||||||| d}|d}|rY||df7}| rm| |df} |jjrm||df}q!| ru| |f} | stdd||| | |fDSt||| | |d S) Nr%)rrrrr7rr6css|] }|dur|VqdSrJr%)rvr%r%r& 8sz&BrosEncoder.forward..)last_hidden_staterrrcross_attentions)r1r enumeraterr$r )r0rrTrrrrrrrrrall_hidden_statesall_self_attentionsall_cross_attentionsnext_decoder_cacherS layer_modulelayer_head_maskr layer_outputsr%r%r&rEsZ       zBrosEncoder.forward) NNNNNNFFT)rrrr,r!rGrr"r$rrr rErHr%r%r2r&rsH       rcr') BrosPoolercs*tt|j|j|_t|_dSrJ)r+r,rrXrmrTanh activationrNr2r%r&r,Ns zBrosPooler.__init__rr5cCs(|dddf}||}||}|S)Nr)rr)r0rfirst_token_tensor pooled_outputr%r%r&rESs  zBrosPooler.forwardrFr%r%r2r&rMs rcr')BrosRelationExtractorcst|j|_|j|_|j|_|j|_t|j|_ t |j|j|j|_ t |j|j|j|_ t td|j|_dS)Nr)r+r, n_relationsrmbackbone_hidden_sizehead_hidden_sizeclassifier_dropout_probrrvdroprXrr Parameterr!r{ dummy_noderNr2r%r&r,]s zBrosRelationExtractor.__init__rrc Cs|||}|jdd|dd}tj||gdd}|||}| |d|d|j |j }| |d|d|j |j }t | dddd| dddd}|S)Nrraxisr6r)rrr unsqueezerepeatr:r!r<rr;r rrr)r0rr dummy_vecrelation_scorer%r%r&rEks zBrosRelationExtractor.forwardrFr%r%r2r&r \s r c@seZdZeZdZddZdS)BrosPreTrainedModelbroscCst|tjr |jjjd|jjd|jdur|jj dSdSt|tj rC|jjjd|jjd|j durA|jj|j  dSdSt|tj rX|jj |jj ddSdS)zInitialize the weightsr))meanstdNg?)rrrXweightdatanormal_r1initializer_rangerVzero_rkr`rtfill_)r0moduler%r%r& _init_weightss      z!BrosPreTrainedModel._init_weightsN)rrrr config_classbase_model_prefixr$r%r%r%r&r~s rc"seZdZdfdd ZddZddZdd Ze dd ee j d ee j d ee j dee j dee j dee j dee j dee j dee j dee e j dee dee dee dee deee j effddZZS) BrosModelTcsNt|||_t||_t||_t||_|rt |nd|_ | dS)zv add_pooling_layer (bool, *optional*, defaults to `True`): Whether to add a pooling layer N) r+r,r1r_rrUbbox_embeddingsrencoderrpooler init_weights)r0r1add_pooling_layerr2r%r&r,s     zBrosModel.__init__cCs|jjSrJrro)r0r%r%r&get_input_embeddingsszBrosModel.get_input_embeddingscCs ||j_dSrJr-)r0rr%r%r&set_input_embeddingss zBrosModel.set_input_embeddingscCs*|D]\}}|jj|j|qdS)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)itemsr)rrr)r0heads_to_prunerrr%r%r& _prune_headsszBrosModel._prune_headsNr}rOrrfrerr~rrrrrrrr5c" Cs| dur| n|jj} | dur| n|jj} |dur|n|jj}|jjr-| dur(| n|jj} nd} |dur;|dur;td|durD|}n|durQ|dd}ntd|dur]td|\}}|durh|jn|j}| durx| ddj dnd}|durt j ||d }|durt |j d r|j jddd|f}|||}|}n t j|t j|d }||||}|jjr|dur|\}}}||f}| durt j ||d } || }nd}|||jj}|j |||||d }|j dd kr|ddddgdf}||jj}||}|j||||||| | | | |d }|d} |jdur+|| nd}!|s:| |!f|ddSt| |!|j|j|j|jdS)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 ```NFzDYou cannot specify both input_ids and inputs_embeds at the same timer7z5You have to specify either input_ids or inputs_embedszYou have to specify bboxrr6)rirfrg)r}rerfr~r)rrr6rr6rrr) rTrrrrrrrrrr)r pooler_outputrrrr)r1rruse_return_dictrrrr:rirr!onesrrrfrzr{r|get_extended_attention_maskinvert_attention_mask get_head_maskr bbox_scaler(r)r*r rrrr)"r0r}rOrrfrerr~rrrrrrrrrrrirrrextended_attention_maskencoder_batch_sizeencoder_sequence_lengthrencoder_hidden_shapeencoder_extended_attention_maskembedding_output scaled_bboxbbox_position_embeddingsencoder_outputssequence_outputr r%r%r&rEs)        zBrosModel.forward)T)NNNNNNNNNNNNNN)rrrr,r.r/r2rrr!rGlistr"rrr$r rErHr%r%r2r&r'sf     r'ceZdZdgZfddZe            ddeejdeejdeejdeejd eejd eejd eejd eejd eejdee dee dee de e eje ffddZ ZS)BrosForTokenClassificationr*cs^t||j|_t||_t|dr|jn|j}t ||_ t |j |j|_ |dSNclassifier_dropout)r+r, num_labelsr'rrrIrwrrvrxrXrm classifierr+r0r1rIr2r%r&r,Js    z#BrosForTokenClassification.__init__Nr}rOrbbox_first_token_maskrfrerr~labelsrrrr5c  Cs| dur| n|jj} |j|||||||| | | d } | d}||}||}d}| durXt}|durK|d}||d|j|| d|}n ||d|j| d}| sn|f| dd}|durl|f|S|St||| j | j dS)a bbox ('torch.FloatTensor' of shape '(batch_size, num_boxes, 4)'): Bounding box coordinates for each token in the input sequence. Each bounding box is a list of four values (x1, y1, x2, y2), where (x1, y1) is the top left corner, and (x2, y2) is the bottom right corner of the bounding box. bbox_first_token_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to indicate the first token of each bounding box. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. Examples: ```python >>> import torch >>> from transformers import BrosProcessor, BrosForTokenClassification >>> processor = BrosProcessor.from_pretrained("jinho8345/bros-base-uncased") >>> model = BrosForTokenClassification.from_pretrained("jinho8345/bros-base-uncased") >>> encoding = processor("Hello, my dog is cute", add_special_tokens=False, return_tensors="pt") >>> bbox = torch.tensor([[[0, 0, 1, 1]]]).repeat(1, encoding["input_ids"].shape[-1], 1) >>> encoding["bbox"] = bbox >>> outputs = model(**encoding) ```N) rOrrfrerr~rrrrr7r6rlogitsrr) r1r5rrxrKrr;rJr rr)r0r}rOrrMrfrerr~rNrrrrrDrPrloss_fctrr%r%r&rEWsD,   z"BrosForTokenClassification.forward NNNNNNNNNNNNrrr"_keys_to_ignore_on_load_unexpectedr,rrr!rGrrr$r rErHr%r%r2r&rGFsV       rGa 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 seZdZdgZfddZe             ddeejdeejdeejdeejd eejd eejd eejd eejd eejdeejdee dee dee de e eje ffddZ ZS)!BrosSpadeEEForTokenClassificationr*c st|||_|j|_|j|_|j|_t||_t |dr"|j n|j }t t |t |j|jt |t |j|j|_t||_|dSrH)r+r,r1rJr rmr r'rrrIrwr SequentialrvrXinitial_token_classifierr subsequent_token_classifierr+rLr2r%r&r,s    z*BrosSpadeEEForTokenClassification.__init__Nr}rOrrMrfrerr~initial_token_labelssubsequent_token_labelsrrrr5c Cs| dur| n|jj} |j|||||||| | | d }|d}|dd}||dd}|||d}d|}|j\}}|j }t j |t |dg |gdd}||dddddft |jj}t ||dj |t jd}||dddddft |jj}|d}d}| dur| durt}| d} |dur|d}||d|j|| |}n ||d|j| }| d} ||d|d|| |}||}| s||f|dd}|dur|f|S|St||||j|jd 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) ```N r}rOrrfrerr~rrrrrrrirhr7r6)rrrrr)r1r5rr\rrWrXsqueezerrir!r<r{rr masked_fillfinforhmineyer;rrJrrr)r0r}rOrrMrfrerr~rYrZrrrrlast_hidden_statesrrinv_attention_maskrmax_seq_lengthriinvalid_token_maskself_token_masksubsequent_token_maskrrQinitial_token_losssubsequent_token_lossrr%r%r&rEsp1  &     z)BrosSpadeEEForTokenClassification.forward) NNNNNNNNNNNNN)rrrrTr,rrr!rGrrr$rrErHr%r%r2r&rUs\      rUz Bros Model with a token classification head on top (a entity_linker layer on top of the hidden-states output) e.g. for Entity-Linking. The entity_linker is used to predict intra-entity links (one entity to another entity). crF)!BrosSpadeELForTokenClassificationr*csxt|||_|j|_|j|_|j|_t||_t |dr"|j n|j t ||_ |dSt ||_ |dSrH)r+r,r1rJr rmr r'rrrIrwr  entity_linkerr+rNr2r%r&r,Ss      z*BrosSpadeELForTokenClassification.__init__Nr}rOrrMrfrerr~rNrrrr5c  Cs~| dur| n|jj} |j|||||||| | | d } | d}|dd}|||d}d}| durt}|j\}}|j }t ||dj |t j d}|d}t j|t j|dgt j |dgdd}||dddddft |jj}||dddddft |jj}||d|d|| d|}| s|f| d d}|dur|f|S|St||| j| jd S) a bbox ('torch.FloatTensor' of shape '(batch_size, num_boxes, 4)'): Bounding box coordinates for each token in the input sequence. Each bounding box is a list of four values (x1, y1, x2, y2), where (x1, y1) is the top left corner, and (x2, y2) is the bottom right corner of the bounding box. bbox_first_token_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to indicate the first token of each bounding box. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. Examples: ```python >>> import torch >>> from transformers import BrosProcessor, BrosSpadeELForTokenClassification >>> processor = BrosProcessor.from_pretrained("jinho8345/bros-base-uncased") >>> model = BrosSpadeELForTokenClassification.from_pretrained("jinho8345/bros-base-uncased") >>> encoding = processor("Hello, my dog is cute", add_special_tokens=False, return_tensors="pt") >>> bbox = torch.tensor([[[0, 0, 1, 1]]]).repeat(1, encoding["input_ids"].shape[-1], 1) >>> encoding["bbox"] = bbox >>> outputs = model(**encoding) ```Nr[rrr\r7rgrr6rO)r1r5rr\rrkr]rrrir!rarrr;r<r{r^r_rhr`r rr)r0r}rOrrMrfrerr~rNrrrrrbrPrrQrrdrirfmaskrr%r%r&rEasR+   (($z)BrosSpadeELForTokenClassification.forwardrRrSr%r%r2r&rjJsV      rj)rr'rGrUrj)8r r dataclassesrtypingrrr!torch.utils.checkpointrtorch.nnr activationsrmodeling_layersr modeling_outputsr r r modeling_utilsr pytorch_utilsrrrutilsrrrconfiguration_brosr get_loggerrloggerrModuler(rIrUr_rrrrrrrrr rr'rGrUrj__all__r%r%r%r&sh        B 7ZL"2g  r