+ j"^RIHtHt^RIHt^RIHt^RIHtH t ^RI H t H t H t ^RIHtHt^RIHtHtHt]!4'd^RIt]P,!]4t!R R ] R R 7t!R R]4tR.tR#))OptionalUnion)PaliGemmaProcessor) BatchFeature) ImageInputmake_flat_list_of_images)ProcessingKwargsProcessorMixinUnpack)PreTokenizedInput TextInput)auto_docstringis_torch_availableloggingNc>]tRt^ tRRRRRRR/RRR R R /R R R//tRtR#)ColPaliProcessorKwargs text_kwargspaddinglongestreturn_mm_token_type_idsFreturn_text_replacement_offsets images_kwargs data_formatchannels_firstdo_convert_rgbT common_kwargsreturn_tensorsptN)__name__ __module__ __qualname____firstlineno__ _defaults__static_attributes__rl/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/colpali/modular_colpali.pyrr sB y & -u  + d  *D1 Ir&rF)totalc"aa]tRt^/to]tRV3RlV3Rlllt]V3RlRl4t] RV3RlRll4t RV3RlRllt RR lt V3R lR lt ] !R R 7RV3RlRll4t] !RR 7V3RlRl4tRV3RlRlltRtVtV;t#)ColPaliProcessorc&<V^8dQhRS[RS[/#)visual_prompt_prefix query_prefixstr)format __classdict__s"r' __annotate__ColPaliProcessor.__annotate__2s%ll " l  lr&c B<W@nWPn\SV` WVR7R#)z visual_prompt_prefix (`str`, *optional*, defaults to `"Describe the image."`): A string that gets tokenized and prepended to the image tokens. query_prefix (`str`, *optional*, defaults to `"Question: "`): A prefix to be used for the query. )image_processor tokenizer chat_templateN)r-r.super__init__)selfr6r7r8r-r. __class__s&&&&&&r'r:ColPaliProcessor.__init__2s$%9!( ]jkr&c <V^8dQhRS[/#)r,returnr/)r1r2s"r'r3r4Es((#(r&c .VPP#)zr Return the query augmentation token. Query augmentation buffers are used as reasoning buffers during inference. )r7 pad_token)r;s&r'query_augmentation_token)ColPaliProcessor.query_augmentation_tokenDs~~'''r&c<V^8dQhRS[R,RS[S[,S[S[,,S[S[,,RS[S[,RS[/#)r,imagesNtextkwargsr?)rr r listr rr)r1r2s"r'r3r4NsXT!++d9o=EV@WW/0   r&c zVeVe \R4hRVR&VP!VP3RVPP/VBpVR,P RVP ^ ,4pVeyVUu.uF+q`PP VP V V R2NK- ppVR,PRR4e'VR,R;;,VP, uu&\P!RR VR V/VBpVe'VR ,PVR ,^8HR4VR &V#uupi)Nz5Only one of text or images can be processed at a timeTreturn_token_type_idstokenizer_init_kwargsrsuffix max_lengthrErF input_idstoken_type_idslabelsri) ValueError _merge_kwargsvalid_processor_kwargsr7 init_kwargspoprB bos_tokenr.getimage_seq_lengthr __call__ masked_fill)r;rErFrG output_kwargsrLsample model_inputss&&&, r'rZColPaliProcessor.__call__MsL   2TU U*.&'**  ' ' "&.."<"<  }-11(DS>SS:%..YfY4Y=Y  %1+%>%J%J1D8c<V^8dQhRS[R,RS[S[,S[S[,,S[S[,,R,RS[S[,/#)r,rENrFrG)rr r rHr r )r1r2s"r'r3r4isYGGT!G++d9o=EV@WWZ^^G)* Gr&c ^\P!WR7VfVf \R4hR#R#))rErFNz&Either text or images must be provided)r validate_inputsrR)r;rErFrGs&&&,r'rb ColPaliProcessor.validate_inputsis. &&f@ t?X?X>YY[\+ T4'' s:Bc,<V^8dQhRS[RS[RS[/#)r, image_inputs image_idxr?)dictintr0)r1r2s"r'r3r4}s#8888SV8r&c <VPVP,#N)rhrY)r;rlrmrGs&&&,r'replace_image_token$ColPaliProcessor.replace_image_token}s$"7"777r&znThis method forwards the `images` and `kwargs` arguments to ColPaliProcessor's [`~ColPaliProcessor.__call__`].) custom_introc4<V^8dQhRS[R,RS[/#)r,rENr?)rr)r1r2s"r'r3r4s$66T!6  6r&c *VP!RRV/VB#)rErrZ)r;rErGs&&,r'process_imagesColPaliProcessor.process_imagess}}5F5f55r&zlThis method forwards the `text` and `kwargs` arguments to ColPaliProcessor's [`~ColPaliProcessor.__call__`].cF<V^8dQhRS[S[S[,,RS[/#)r,rFr?)r rHr)r1r2s"r'r3r4s(22$y/)2  2r&c *VP!RRV/VB#)rFrrw)r;rFrGs&&,r'process_queries ColPaliProcessor.process_queriess}}1$1&11r&c <V^8dQhRS[RS[R,3,RS[RS[R,3,RS[RS[R,RS[RS[3,R R/#) r,query_embeddingsz torch.Tensorpassage_embeddings batch_size output_dtypez torch.dtype output_devicez torch.devicer?)rrHrorr0)r1r2s"r'r3r4st>(>(^0D DE>(".$~2F"FG>( >( }- >( ^S01 >( >(r&c  8\V4^8Xd \R4h\V4^8Xd \R4hV^,PV^,P8wd \R4hV^,PV^,P8wd \R4hVfV^,Pp.p\ ^\V4V4EF1p.p\ P PPPWWs,R^R7p \ ^\V4V4Fp \ P PPPW*W,R^R7p VP\ P!RW4P^R7^,P^R74K VP\ P!V^R7PV4PV44EK4 \ P!V^R7#) a Compute the late-interaction/MaxSim score (ColBERT-like) for the given multi-vector query embeddings (`qs`) and passage embeddings (`ps`). For ColPali, a passage is the image of a document page. Because the embedding tensors are multi-vector and can thus have different shapes, they should be fed as: (1) a list of tensors, where the i-th tensor is of shape (sequence_length_i, embedding_dim) (2) a single tensor of shape (n_passages, max_sequence_length, embedding_dim) -> usually obtained by padding the list of tensors. Args: query_embeddings (`Union[torch.Tensor, list[torch.Tensor]`): Query embeddings. passage_embeddings (`Union[torch.Tensor, list[torch.Tensor]`): Passage embeddings. batch_size (`int`, *optional*, defaults to 128): Batch size for computing scores. output_dtype (`torch.dtype`, *optional*, defaults to `torch.float32`): The dtype of the output tensor. If `None`, the dtype of the input embeddings is used. output_device (`torch.device` or `str`, *optional*, defaults to "cpu"): The device of the output tensor. Returns: `torch.Tensor`: A tensor of shape `(n_queries, n_passages)` containing the scores. The score tensor is saved on the "cpu" device. zNo queries providedzNo passages providedz/Queries and passages must be on the same devicez-Queries and passages must have the same dtypeT) batch_first padding_valuez bnd,csd->bcns)dim)rgrRdevicedtyperftorchnnutilsrnn pad_sequenceappendeinsummaxsumcatto) r;rrrrrscoresi batch_scores batch_queriesjbatch_passagess &&&&&& r'score_retrieval ColPaliProcessor.score_retrievals@  A %23 3 ! "a '34 4 A  % %);A)>)E)E ENO O A  $ $(:1(=(C(C CLM M  +A.44L%'q#./:\]"A"##LL-PTTYZT[\]^bbghbi C MM%))La8;;LILL][ \=yyQ''r&)r.r-)NNNzDescribe the image.z Question: )NNrq)Ncpu)r r!r"r#rrTr:propertyrBrrZrbrerrrxr|rr%__classdictcell__ __classcell__)r<r2s@@r'r*r*/s3ll$((6GG(88F666D22>(>(>(r&r*)typingrr2transformers.models.paligemma.processing_paligemmarfeature_extraction_utilsr image_utilsrrprocessing_utilsr r r tokenization_utils_baser r rrrrr get_loggerr loggerrr*__all__rr&r'rsk#Q4?HHC@@   H % -U c()c(N r&