o i@sZdZddlmZmZddlmZddlmZddlm Z m Z m Z GdddeZ dgZ d S) z Processor class for MarkupLM. )OptionalUnion) TensorType)ProcessorMixin) BatchEncodingPaddingStrategyTruncationStrategyc @seZdZdZdZdZdZ                   dded eee e fd eee e fd e e d e d e e de ede edededededede ee efdefddZdS)MarkupLMProcessoraJ Constructs a MarkupLM processor which combines a MarkupLM feature extractor and a MarkupLM tokenizer into a single processor. [`MarkupLMProcessor`] offers all the functionalities you need to prepare data for the model. It first uses [`MarkupLMFeatureExtractor`] to extract nodes and corresponding xpaths from one or more HTML strings. Next, these are provided to [`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`], which turns them into token-level `input_ids`, `attention_mask`, `token_type_ids`, `xpath_tags_seq` and `xpath_subs_seq`. Args: feature_extractor (`MarkupLMFeatureExtractor`): An instance of [`MarkupLMFeatureExtractor`]. The feature extractor is a required input. tokenizer (`MarkupLMTokenizer` or `MarkupLMTokenizerFast`): An instance of [`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`]. The tokenizer is a required input. parse_html (`bool`, *optional*, defaults to `True`): Whether or not to use `MarkupLMFeatureExtractor` to parse HTML strings into nodes and corresponding xpaths. MarkupLMFeatureExtractor)MarkupLMTokenizerMarkupLMTokenizerFastTNFradd_special_tokenspadding truncation max_lengthstridepad_to_multiple_ofreturn_token_type_idsreturn_attention_maskreturn_overflowing_tokensreturn_special_tokens_maskreturn_offsets_mapping return_lengthverbosereturn_tensorsreturncKs|jr)|dur td|dus|dus|durtd||}|d}|d}n|dur1td|dus9|dur=td|durL|jrLt|trL|g}|jdid|durW|n|d |dur`|n8dd|d |d |d |d |d| d| d| d| d| d|d|d|d|d|d||}|Sd|d |d |d |d |d| d| d| d| d| d|d|d|d|d|d||}|S)a This method first forwards the `html_strings` argument to [`~MarkupLMFeatureExtractor.__call__`]. Next, it passes the `nodes` and `xpaths` along with the additional arguments to [`~MarkupLMTokenizer.__call__`] and returns the output. Optionally, one can also provide a `text` argument which is passed along as first sequence. Please refer to the docstring of the above two methods for more information. NzDMake sure to pass HTML strings in case `parse_html` is set to `True`zUPlease don't pass nodes, xpaths nor node labels in case `parse_html` is set to `True`nodesxpathsz@You have passed HTML strings but `parse_html` is set to `False`.zIMake sure to pass nodes and xpaths in case `parse_html` is set to `False`text text_pair node_labelsrrrrrrrrrrrrrr) parse_html ValueErrorfeature_extractor isinstancestr tokenizer)self html_stringsrrr! questionsrrrrrrrrrrrrrrkwargsfeaturesencoded_inputsr"r"m/home/ubuntu/veenaModal/venv/lib/python3.10/site-packages/transformers/models/markuplm/processing_markuplm.py__call__2s!              zMarkupLMProcessor.__call__)NNNNNTFNNrNNNFFFFTN)__name__ __module__ __qualname____doc__feature_extractor_classtokenizer_classr#boolrr'rr rintrrr0r"r"r"r/r sn      r N)r4typingrr file_utilsrprocessing_utilsrtokenization_utils_baserrr r __all__r"r"r"r/s   i