§ ‚ŠtjÿãóB—dZddlmZddlZddlmZddlmZddlmZddl m Z dd l m Z dd l mZmZmZmZd d lmZeGd „de ¦«¦«Zed¬¦«eGd„de¦«¦«¦«Zed¬¦«Gd„de¦«¦«ZddgZdS)zPyTorch ColPali modelé)Ú dataclassN)Únn)Ú AutoModelé)ÚCache)ÚPreTrainedModel)ÚUnpack)Ú ModelOutputÚTransformersKwargsÚauto_docstringÚcan_return_tupleé)Ú ColPaliConfigcól‡—eZdZUeed<dZdZgZdZdZ dZ e j ¦«ˆfd„¦«Z ˆxZS)ÚColPaliPreTrainedModelÚconfigÚmodel)ÚimageÚtextTcóJ•—t¦« |¦«dS©N)ÚsuperÚ _init_weights)ÚselfÚmoduleÚ __class__s €új/var/www/html/CA-Chatbot/venv/lib/python3.11/site-packages/transformers/models/colpali/modeling_colpali.pyrz$ColPaliPreTrainedModel._init_weights(s!ø€å ‰Œ×Ò˜fÑ%Ô%Ð%Ð%Ð%ó)Ú__name__Ú __module__Ú __qualname__rÚ__annotations__Úbase_model_prefixÚinput_modalitiesÚ_no_split_modulesÚ_supports_sdpaÚ_supports_flash_attnÚ_supports_flex_attnÚtorchÚno_gradrÚ __classcell__©rs@rrrs{ø€€€€€€à ÐÐÑØÐØ(ÐØÐØ€NØÐØÐà€U„]_„_ð&ð&ð&ð&ñ„_ð&ð&ð&ð&ð&rrz3 Base class for ColPali embeddings output. )Ú custom_introcóÖ—eZdZUdZdZejdzed<dZej dzed<dZ e dzed<dZ e ejdzed<dZe ejdzed<dZejdzed<dS) ÚColPaliForRetrievalOutputaä loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction). embeddings (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`): The embeddings of the model. past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): It is a [`~cache_utils.Cache`] instance. For more details, see our [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache). Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. image_hidden_states (`torch.FloatTensor`, *optional*): A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder after projecting last hidden state. NÚlossÚ embeddingsÚpast_key_valuesÚ hidden_statesÚ attentionsÚimage_hidden_states)rr r!Ú__doc__r0r)Ú FloatTensorr"r1ÚTensorr2rr3Útupler4r5©rrr/r/-s´€€€€€€ð ð ð&*€Dˆ%Ô ˜dÑ "Ð)Ð)Ñ)Ø&*€J” ˜tÑ#Ð*Ð*Ñ*Ø$(€OU˜T‘\Ð(Ð(Ñ(Ø59€M5˜Ô*Ô+¨dÑ2Ð9Ð9Ñ9Ø26€JeÔ'Ô(¨4Ñ/Ð6Ð6Ñ6Ø48ИÔ*¨TÑ1Ð8Ð8Ñ8Ð8Ð8rr/u/ The ColPali architecture leverages VLMs to construct efficient multi-vector embeddings directly from document images (“screenshotsâ€) for document retrieval. The model is trained to maximize the similarity between these document embeddings and the corresponding query embeddings, using the late interaction method introduced in ColBERT. Using ColPali removes the need for potentially complex and brittle layout recognition and OCR pipelines with a single model that can take into account both the textual and visual content (layout, charts, etc.) of a document. ColPali is part of the ColVision model family, which was first introduced in the following paper: [*ColPali: Efficient Document Retrieval with Vision Language Models*](https://huggingface.co/papers/2407.01449). c 󨇗eZdZdZdefˆfd„ Zee d dej dzdej dzdej dzde e d ef d „¦«¦«ZˆxZS) ÚColPaliForRetrievalÚvlmrcóp•—t¦« |¦«||_|jjj|_t j|j¦«|_|jj |_ tj |jjjj |j ¦«|_ | ¦«dSr)rÚ__init__rÚ vlm_configÚ text_configÚ vocab_sizerÚ from_configr=Ú embedding_dimrÚLinearÚ hidden_sizeÚembedding_proj_layerÚ post_init)rrrs €rr?zColPaliForRetrieval.__init__\s”ø€Ý ‰Œ×Ò˜Ñ Ô Ð ØˆŒ Ø Ô+Ô7ÔBˆŒåÔ(¨Ô):Ñ;Ô;ˆŒà!œ[Ô6ˆÔÝ$&¤IØ ŒKÔ "Ô .Ô :Ø Ô ñ% ô% ˆÔ!ð ŠÑÔÐÐÐrNÚ input_idsÚ pixel_valuesÚattention_maskÚkwargsÚreturnc óú—|| |j¬¦«}| dd¦«}|€ |jj}|jd |||ddœ|¤Ž}|r|jnd}||jnd}|d} |jj j} | |  | ¦«¦«} | |   dd¬¦«z } || |  d¦«z} t| |j ||j|¬¦«S) N)ÚdtypeÚoutput_hidden_statesT)rIrKrJrPréÿÿÿÿ)ÚdimÚkeepdim)r1r2r3r4r5r:)ÚtorOÚpoprrPr=r3r5rGÚweightÚnormÚ unsqueezer/r2r4) rrIrJrKrLrPÚ vlm_outputÚvlm_hidden_statesÚvlm_image_hidden_statesÚlast_hidden_statesÚ proj_dtyper1s rÚforwardzColPaliForRetrieval.forwardksC€ð Ð #Ø'Ÿ?š?°´˜?Ñ<Ô<ˆLØ%ŸzšzÐ*@À$ÑGÔGÐØ Ð 'Ø#'¤;Ô#CÐ àT”Xð ØØ)Ø%Ø!%ð  ð ð ð  ð ˆ ð9MÐV˜JÔ4Ð4ÐRVÐØDPÐD\ *Ô"@Ð"@ÐbfÐà'¨œ]ÐØÔ.Ô5Ô;ˆ Ø×.Ò.Ð/A×/DÒ/DÀZÑ/PÔ/PÑQÔQˆ ð  *§/¢/°bÀ$ /Ñ"GÔ"GÑGˆ à Ð %Ø# n×&>Ò&>¸rÑ&BÔ&BÑBˆJå(Ø!Ø&Ô6Ø+Ø!Ô,Ø 7ð  ñ ô ð r)NNN)rr r!r#rr?r r r)Ú LongTensorr7r8r r r/r^r+r,s@rr<r<KsÔø€€€€€ðÐð ˜}ð ð ð ð ð ð ðØð.2Ø15Ø.2ð ' ð' àÔ# dÑ*ð' ðÔ'¨$Ñ.ð' 𜠠tÑ+ð ' ð Ð+Ô,ð ' ð #ð ' ð' ð' ñ„^ñÔð' ð' ð' ð' ð' rr<)r6Ú dataclassesrr)rÚ transformersrÚ cache_utilsrÚmodeling_utilsrÚprocessing_utilsr Úutilsr r r r Úconfiguration_colpalirrr/r<Ú__all__r:rrúrhs­ððÐà!Ð!Ð!Ð!Ð!Ð!à € € € ØÐÐÐÐÐà"Ð"Ð"Ð"Ð"Ð"à Ð Ð Ð Ð Ð Ø-Ð-Ð-Ð-Ð-Ð-Ø&Ð&Ð&Ð&Ð&Ð&ØVÐVÐVÐVÐVÐVÐVÐVÐVÐVÐVÐVØ0Ð0Ð0Ð0Ð0Ð0ðð &ð &ð &ð &ð &˜_ñ &ô &ñ„ð &ð€ððñôð  ð9ð9ð9ð9ð9  ñ9ô9ñ „ñ ôð 9ð0€ð ð ñ ô ð; ð; ð; ð; ð; Ð0ñ; ô; ñ ô ð; ð~Øð €€€r