qidZddlmZddlmZddlmZddlmZm Z m Z m Z ddl m Z mZddlmZej"eZGd d e d ZGd ded ZeGdde ZdgZy)z Processor class for UDOP. )logging) BatchFeature) ImageInput)ProcessingKwargsProcessorMixin TextKwargsUnpack)PreTokenizedInput TextInput)auto_docstringcjeZdZUeeeeezdzed<eeeeeeezdzed<y)UdopTextKwargsN word_labelsboxes)__name__ __module__ __qualname__listint__annotations__Z/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/udop/processing_udop.pyrrsCcT$s)_,t33 S ?T$tCy/2 2T 99rrF)totalc 4eZdZUeed<ddddddddddd iZy)UdopProcessorKwargs text_kwargsTFr) add_special_tokenspadding truncationstridereturn_overflowing_tokensreturn_special_tokens_maskreturn_offsets_mapping return_lengthverboseN)rrrrr _defaultsrrrrr#s2"&).*/&+"   Irrc eZdZdZfdZe d dedzdeeze eze ezde e de fdZ d Zed ZxZS) UdopProcessoras Constructs a UDOP processor which combines a LayoutLMv3 image processor and a UDOP tokenizer into a single processor. [`UdopProcessor`] offers all the functionalities you need to prepare data for the model. It first uses [`LayoutLMv3ImageProcessor`] to resize, rescale and normalize document images, and optionally applies OCR to get words and normalized bounding boxes. These are then provided to [`UdopTokenizer`], which turns the words and bounding boxes into token-level `input_ids`, `attention_mask`, `token_type_ids`, `bbox`. Optionally, one can provide integer `word_labels`, which are turned into token-level `labels` for token classification tasks (such as FUNSD, CORD). Additionally, it also supports passing `text_target` and `text_pair_target` to the tokenizer, which can be used to prepare labels for language modeling tasks. c&t|||y)N)super__init__)selfimage_processor tokenizer __class__s rr-zUdopProcessor.__init__Es )4rNimagestextkwargsreturnc 6|jtfd|jji|}|dj dd}|dj dd}|dj dd}|dj dd}|dj dd} |dj d d} |j jr | td |j jr | td |r | s td | |jdi|dS|j dd |i|d} | j dd} | j dd} |dj d d|dj dd||dd<||n| |dd<||dd<|3|j jr|t|tr|g}| |dd<|jdd||n| i|d}|dur|j| d|d| d<| j|| S)Ntokenizer_init_kwargsrrr text_pairr#Fr% text_targetzdYou cannot provide bounding boxes if you initialized the image processor with apply_ocr set to True.zaYou cannot provide word labels if you initialized the image processor with apply_ocr set to True.zKYou cannot return overflowing tokens without returning the offsets mapping.r2 images_kwargswordstext_pair_targetr3T pixel_valuesoverflow_to_sample_mappingr) _merge_kwargsrr0 init_kwargspopgetr/ apply_ocr ValueError isinstancestrget_overflowing_imagesupdate)r.r2r3r4 output_kwargsrrr8r#r%r9featuresfeatures_wordsfeatures_boxesencoded_inputss r__call__zUdopProcessor.__call__Hs+**  "&.."<"<  m,00$?#M266}dK !-044[$G $1-$@$D$DE`bg$h!!.}!=!A!ABZ\a!b#M266}dK    ) )u/@v     ) ){/Fs  %-Cjk k  "!4>> . ,t++\6\]?=[\H%\\'48N%\\'48N - ( , ,]D A - ( , ,-? F8AM- ( 5=B=NETbM- ( 1:EM- ( 7D$8$8$B$ByGXdC( 6D .N )D0+/+F+F^,n=Y.Z,( OON +Orcg}|D]}|j||t|t|k7r#tdt|dt||S)Nz`Expected length of images to be the same as the length of `overflow_to_sample_mapping`, but got z and )appendlenrD)r.r2r>images_with_overflow sample_idxs rrGz$UdopProcessor.get_overflowing_imagessy!4 res~!2%TTC#   H %:Zu: *%"pTNpTpTf  r