+ jLU^Rt^RIHt^RIt^RIHtHt^RIHt^RIH t ^RI H t H t H t ^RIHt^RIHt] P$!]4t] !R R 7]!R R ] 444t] !R R 7]!R R] 444t] !R R 7]!RR] 444t] !RR] 44t!RR]4t!RR]4t!RR]4t!RR]4t!RR]4t] !RR 7!RR]44t] !R R 7!R!R"]44t] !R#R 7!R$R%]44t .R&Ot!R#)'z5PyTorch DPR model for Open Domain Question Answering.) dataclassN)Tensornn)BaseModelOutputWithPooling)PreTrainedModel) ModelOutputauto_docstringlogging) BertModel DPRConfigz6 Class for outputs of [`DPRQuestionEncoder`]. ) custom_introc8a]tRt^(toRtRtRtV3RltRtVt R#)DPRContextEncoderOutputa pooler_output (`torch.FloatTensor` of shape `(batch_size, embeddings_size)`): The DPR encoder outputs the *pooler_output* that corresponds to the context representation. Last layer hidden-state of the first token of the sequence (classification token) further processed by a Linear layer. This output is to be used to embed contexts for nearest neighbors queries with questions embeddings. Nc<V^8dQh/S[P;R&S[S[PR3,R,;R&S[S[PR3,R,;R&# pooler_output.N hidden_states attentionstorch FloatTensortuple)format __classdict__s"e/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/dpr/modeling_dpr.py __annotate__$DPRContextEncoderOutput.__annotate__(\$$$**C/047> e'',-4;! __name__ __module__ __qualname____firstlineno____doc__rr__annotate_func____static_attributes____classdictcell__rs@rrr(;?M7;J!r rc8a]tRt^;toRtRtRtV3RltRtVt R#)DPRQuestionEncoderOutputa pooler_output (`torch.FloatTensor` of shape `(batch_size, embeddings_size)`): The DPR encoder outputs the *pooler_output* that corresponds to the question representation. Last layer hidden-state of the first token of the sequence (classification token) further processed by a Linear layer. This output is to be used to embed questions for nearest neighbors queries with context embeddings. Nc<V^8dQh/S[P;R&S[S[PR3,R,;R&S[S[PR3,R,;R&#rr)rrs"rr%DPRQuestionEncoderOutput.__annotate__;rr r!r"r+s@rr.r.;r,r r.c@a]tRt^NtoRtRtRtRtRtV3Rlt Rt Vt R#)DPRReaderOutputa start_logits (`torch.FloatTensor` of shape `(n_passages, sequence_length)`): Logits of the start index of the span for each passage. end_logits (`torch.FloatTensor` of shape `(n_passages, sequence_length)`): Logits of the end index of the span for each passage. relevance_logits (`torch.FloatTensor` of shape `(n_passages, )`): Outputs of the QA classifier of the DPRReader that corresponds to the scores of each passage to answer the question, compared to all the other passages. Nc<V^8dQh/S[P;R&S[PR,;R&S[PR,;R&S[S[PR3,R,;R&S[S[PR3,R,;R&#)r start_logitsN end_logitsrelevance_logits.rrr)rrs"rrDPRReaderOutput.__annotate__Ns"####$!!D(/%&''$.5'(**C/047>)*e'',-4;+r r!) r#r$r%r&r'r5r6rrr(r)r*r+s@rr2r2Ns(,0J15:>M7;J+r r2c]tRt^ftRtRtR#)DPRPreTrainedModelTr!N)r#r$r%r&_supports_sdpar)r!r rr9r9fsNr r9ctaa]tRt^ktoRtV3RlV3RlltR V3RlRllt]V3RlRl4tRt Vt V;t #) DPREncoder bert_modelc <V^8dQhRS[/#rconfigr )rrs"rrDPREncoder.__annotate__ns  y r c<\SV`V4\VRR7VnVPPP ^8:d \ R4hVPVnVP^8dE\P!VPPP VP4Vn VP4R#)F)add_pooling_layerz!Encoder hidden_size can't be zeroN) super__init__r r=r@ hidden_size ValueErrorprojection_dimrLinear encode_proj post_initselfr@ __class__s&&rrEDPREncoder.__init__ns  #FeD ?? ! ! - - 2@A A$33    "!yy)?)?)K)KVMbMbcD  r c<V^8dQhRS[RS[R,RS[R,RS[R,RS[RS[RS[R S[S[S[R 3,,/#) r input_idsattention_maskNtoken_type_ids inputs_embedsoutput_attentionsoutput_hidden_states return_dictreturn.)rboolrr)rrs"rrrAysu" " "  "  " } "  " #" "  $eFCK&8 8" r c VPVVVVVVVR7p V ^,p V R,p VP^8dVPV 4p V'gW3V R,,#\V V V PV P R7#)rQrRrSrTrUrVrWrNN)last_hidden_staterrrNNNr_)r=rHrJrrr) rMrQrRrSrTrUrVrWkwargsoutputssequence_output pooled_outputs &&&&&&&&, rforwardDPREncoder.forwardys//))'/!5#" "!*'0    " ,,];M#3gbkA A)-'!//))   r c <V^8dQhRS[/#)rrX)int)rrs"rrrAs222r cVP^8dVPP#VPPP #)r`)rHrJ out_featuresr=r@rF)rMs&rembeddings_sizeDPREncoder.embeddings_sizes8    "##00 0%%111r )r=rJrH)NNNFFF) r#r$r%r&base_model_prefixrErepropertyrkr)r* __classcell__rNrs@@rr<r<ks9$  " " H222r r<cXaa]tRt^toRtV3RlV3RlltRV3RlRlltRtVtV;t #)DPRSpanPredictorencoderc <V^8dQhRS[/#r?r )rrs"rrDPRSpanPredictor.__annotate__syr c&<\SV`V4\V4Vn\P !VPP ^4Vn\P !VPP ^4VnVP4R#)rN) rDrEr<rsrrIrk qa_outputs qa_classifierrKrLs&&rrEDPRSpanPredictor.__init__s_  !&) ))DLL$@$@!DYYt||'C'CQG r cv<V^8dQhRS[RS[RS[R,RS[RS[RS[RS[S[S[R 3,,/#) rrQrRrTNrUrVrWrX.rrYr2r)rrs"rrrusc, , , , } ,  , # , ,  5- -, r c vVeVP4MVP4R,wrVPVVVVVVR7p V ^,p VPV 4p V P^RR7wrV P R4P 4p VP R4P 4pVP V R,4pV PW4p VPW4pVPV4pV'gWV3V R,,#\V VVV PV PR7#)N:NrN)rRrTrUrVrW)dimr\)r4r5r6rrr^) sizersrwsplitsqueeze contiguousrxviewr2rr)rMrQrRrTrUrVrWra n_passagessequence_lengthrbrclogitsr4r5r6s&&&&&&&, rreDPRSpanPredictor.forwards9;D:Oinn&6UbUgUgUijlUm# ,, )'/!5#  "!*1#)<<r<#: #++B/::< ''+668 --og.FG$((E __ZA +00< .>?'"+M M%!-!//))   r )rsrxrw)NFFF) r#r$r%r&rmrErer)r*rorps@@rrrrrs$!, , , r rrc4a]tRt^toRtRtV3RltRtVtR#)DPRPretrainedContextEncoderz An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. ctx_encoderc&<V^8dQh/S[;R&#r?r )rrs"rr(DPRPretrainedContextEncoder.__annotate__  r r!N r#r$r%r&r'rmr(r)r*r+s@rrrs &r rc4a]tRt^toRtRtV3RltRtVtR#)DPRPretrainedQuestionEncoderrquestion_encoderc&<V^8dQh/S[;R&#r?r )rrs"rr)DPRPretrainedQuestionEncoder.__annotate__rr r!Nrr+s@rrrs +r rc4a]tRt^toRtRtV3RltRtVtR#)DPRPretrainedReaderrspan_predictorc&<V^8dQh/S[;R&#r?r )rrs"rr DPRPretrainedReader.__annotate__rr r!Nrr+s@rrrs )r rzf The bare DPRContextEncoder transformer outputting pooler outputs as context representations. c^aa]tRtRtoV3RlV3Rllt]RV3RlRll4tRtVtV;t #)DPRContextEncoderic <V^8dQhRS[/#r?r )rrs"rrDPRContextEncoder.__annotate__ yr cr<\SV`V4Wn\V4VnVP 4R#N)rDrEr@r<rrKrLs&&rrEDPRContextEncoder.__init__ s,   %f- r c<V^8dQhRS[R,RS[R,RS[R,RS[R,RS[R,RS[R,RS[R,R S[S[S[R 3,,/# rrQNrRrSrTrUrVrWrX.)rrYrr)rrs"rrrsY Y D=Y  Y  Y } Y  $; Y #TkY D[Y  !5#5 5Y r c VeTMVPPpVeTMVPPpVeTMVPPpVeVe \ R4hVeVP 4p M#VeVP 4RR p M \ R4hVe VP M VP p Vf4Vf\P!WR7MWPP8gpVf(\P!V \PV R7pVPVVVVVVVR7p V'g V R,#\V PV PV P R7#) a input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary. To match pretraining, DPR input sequence should be formatted with [CLS] and [SEP] tokens as follows: (a) For sequence pairs (for a pair title+text for example): ``` tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP] token_type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1 ``` (b) For single sequences (for a question for example): ``` tokens: [CLS] the dog is hairy . [SEP] token_type_ids: 0 0 0 0 0 0 0 ``` DPR is a model with absolute position embeddings so it's usually advised to pad the inputs on the right rather than the left. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) Examples: ```python >>> from transformers import DPRContextEncoder, DPRContextEncoderTokenizer >>> tokenizer = DPRContextEncoderTokenizer.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base") >>> model = DPRContextEncoder.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base") >>> input_ids = tokenizer("Hello, is my dog cute ?", return_tensors="pt")["input_ids"] >>> embeddings = model(input_ids).pooler_output ```NDYou cannot specify both input_ids and inputs_embeds at the same time5You have to specify either input_ids or inputs_embedsdevicedtyperr[NNrrrr~)r@rUrVrWrGrrrones pad_token_idzeroslongrrrrr rMrQrRrSrTrUrVrWra input_shaperrbs &&&&&&&&, rreDPRContextEncoder.forwardsjd2C1N-TXT_T_TqTq$8$D $++JjJj &1%cd d  "#..*K  &',,.s3KTU U%.%:!! @T@T  !$ ;6;;#;#;;   !"[[EJJvVN""))'/!5## 2; &!//w?T?Tahasas  r )r@rNNNNNNN r#r$r%r&rErrer)r*rorps@@rrrs1 Y Y Y Y r rzh The bare DPRQuestionEncoder transformer outputting pooler outputs as question representations. c^aa]tRtRtoV3RlV3Rllt]RV3RlRll4tRtVtV;t #)DPRQuestionEncoderipc <V^8dQhRS[/#r?r )rrs"rrDPRQuestionEncoder.__annotate__vrr cr<\SV`V4Wn\V4VnVP 4R#r)rDrEr@r<rrKrLs&&rrEDPRQuestionEncoder.__init__vs,   *6 2 r c<V^8dQhRS[R,RS[R,RS[R,RS[R,RS[R,RS[R,RS[R,R S[S[S[R 3,,/#r)rrYr.r)rrs"rrr~sZ Z D=Z  Z  Z } Z  $; Z #TkZ D[Z  "E&#+$6 6Z r c VeTMVPPpVeTMVPPpVeTMVPPpVeVe \ R4hVe#VP W4VP 4p M#VeVP 4RR p M \ R4hVe VPM VPp Vf4Vf\P!WR7MWPP8gpVf(\P!V \PV R7pVPVVVVVVVR7p V'g V R,#\V PV P V P"R7#) a input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary. To match pretraining, DPR input sequence should be formatted with [CLS] and [SEP] tokens as follows: (a) For sequence pairs (for a pair title+text for example): ``` tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP] token_type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1 ``` (b) For single sequences (for a question for example): ``` tokens: [CLS] the dog is hairy . [SEP] token_type_ids: 0 0 0 0 0 0 0 ``` DPR is a model with absolute position embeddings so it's usually advised to pad the inputs on the right rather than the left. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) Examples: ```python >>> from transformers import DPRQuestionEncoder, DPRQuestionEncoderTokenizer >>> tokenizer = DPRQuestionEncoderTokenizer.from_pretrained("facebook/dpr-question_encoder-single-nq-base") >>> model = DPRQuestionEncoder.from_pretrained("facebook/dpr-question_encoder-single-nq-base") >>> input_ids = tokenizer("Hello, is my dog cute ?", return_tensors="pt")["input_ids"] >>> embeddings = model(input_ids).pooler_output ``` Nrrrrr[rrr~)r@rUrVrWrG%warn_if_padding_and_no_attention_maskrrrrrrrrr.rrrrs &&&&&&&&, rreDPRQuestionEncoder.forward}sxd2C1N-TXT_T_TqTq$8$D $++JjJj &1%cd d  "  6 6y Q#..*K  &',,.s3KTU U%.%:!! @T@T  !$ ;6;;#;#;;   !"[[EJJvVN''))'/!5#( 2; '!//w?T?Tahasas  r )r@rrrrps@@rrrps1 Z Z Z Z r rzE The bare DPRReader transformer outputting span predictions. c^aa]tRtRtoV3RlV3Rllt]RV3RlRll4tRtVtV;t #) DPRReaderic <V^8dQhRS[/#r?r )rrs"rrDPRReader.__annotate__rr cr<\SV`V4Wn\V4VnVP 4R#r)rDrEr@rrrrKrLs&&rrEDPRReader.__init__s,   .v6 r c<V^8dQhRS[R,RS[R,RS[R,RS[R,RS[R,RS[R,RS[S[S[R 3,,/#) rrQNrRrTrUrVrWrX.r{)rrs"rrrsL L D=L  L } L  $; L #Tk L D[L  5- -L r c VeTMVPPpVeTMVPPpVeTMVPPpVeVe \ R4hVe#VP W4VP 4pM#VeVP 4RRpM \ R4hVe VPM VPp Vf\P!WR7pVPVVVVVVR7#)a input_ids (`tuple[torch.LongTensor]` of shapes `(n_passages, sequence_length)`): Indices of input sequence tokens in the vocabulary. It has to be a sequence triplet with 1) the question and 2) the passages titles and 3) the passages texts To match pretraining, DPR `input_ids` sequence should be formatted with [CLS] and [SEP] with the format: `[CLS] [SEP] [SEP] ` DPR is a model with absolute position embeddings so it's usually advised to pad the inputs on the right rather than the left. Indices can be obtained using [`DPRReaderTokenizer`]. See this class documentation for more details. [What are input IDs?](../glossary#input-ids) inputs_embeds (`torch.FloatTensor` of shape `(n_passages, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert `input_ids` indices into associated vectors than the model's internal embedding lookup matrix. Examples: ```python >>> from transformers import DPRReader, DPRReaderTokenizer >>> tokenizer = DPRReaderTokenizer.from_pretrained("facebook/dpr-reader-single-nq-base") >>> model = DPRReader.from_pretrained("facebook/dpr-reader-single-nq-base") >>> encoded_inputs = tokenizer( ... questions=["What is love ?"], ... titles=["Haddaway"], ... texts=["'What Is Love' is a song recorded by the artist Haddaway"], ... return_tensors="pt", ... ) >>> outputs = model(**encoded_inputs) >>> start_logits = outputs.start_logits >>> end_logits = outputs.end_logits >>> relevance_logits = outputs.relevance_logits ``` Nrrr)rTrUrVrWr~) r@rUrVrWrGrrrrrr) rMrQrRrTrUrVrWrarrs &&&&&&&, rreDPRReader.forwardsb2C1N-TXT_T_TqTq$8$D $++JjJj &1%cd d  "  6 6y Q#..*K  &',,.s3KTU U%.%:!! @T@T  !"ZZ CN""  '/!5# #  r )r@r)NNNNNNrrps@@rrrs1 L L L L r r)rrr9rrrr)"r' dataclassesrrrrmodeling_outputsrmodeling_utilsrutilsrrr bert.modeling_bertr configuration_dprr get_loggerr#loggerrr.r2r9r<rrrrrrrr__all__r!r rrs<! :- +(   H %