U &cr@sdZddlZddlZddlZddlZddlmZmZddlm Z ddl m Z m Z e eZddiZdd d d d d ddddddddddddddiZdddddddddddddddddddZddiddiddiddiddiddiddiddiddiddiddiddiddiddiddiddiddiddidZd d!Zd"d#ZGd$d%d%e ZGd&d'd'eZGd(d)d)eZd*d+Zd,d-Zd.d/ZGd0d1d1e ZdS)2zTokenization classes.N)ListOptional)BertWordPieceTokenizer)PreTrainedTokenizerPreTrainedTokenizerFast vocab_filez vocab.txtzOhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-vocab.txtzPhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-large-uncased-vocab.txtzMhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-base-cased-vocab.txtzNhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-large-cased-vocab.txtz\https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-multilingual-uncased-vocab.txtzZhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-base-multilingual-cased-vocab.txtzOhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-base-chinese-vocab.txtzfhttps://int-deepset-models-bert.s3.eu-central-1.amazonaws.com/pytorch/bert-base-german-cased-vocab.txtzchttps://s3.amazonaws.com/models.huggingface.co/bert/bert-large-uncased-whole-word-masking-vocab.txtzahttps://s3.amazonaws.com/models.huggingface.co/bert/bert-large-cased-whole-word-masking-vocab.txtzshttps://s3.amazonaws.com/models.huggingface.co/bert/bert-large-uncased-whole-word-masking-finetuned-squad-vocab.txtzqhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-large-cased-whole-word-masking-finetuned-squad-vocab.txtz\https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-cased-finetuned-mrpc-vocab.txtzZhttps://s3.amazonaws.com/models.huggingface.co/bert/bert-base-german-dbmdz-cased-vocab.txtz\https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-german-dbmdz-uncased-vocab.txtzahttps://s3.amazonaws.com/models.huggingface.co/bert/TurkuNLP/bert-base-finnish-cased-v1/vocab.txtzchttps://s3.amazonaws.com/models.huggingface.co/bert/TurkuNLP/bert-base-finnish-uncased-v1/vocab.txtz\https://s3.amazonaws.com/models.huggingface.co/bert/wietsedv/bert-base-dutch-cased/vocab.txt)zbert-base-uncasedzbert-large-uncasedzbert-base-casedzbert-large-casedzbert-base-multilingual-uncasedzbert-base-multilingual-casedzbert-base-chinesezbert-base-german-casedz%bert-large-uncased-whole-word-maskingz#bert-large-cased-whole-word-maskingz5bert-large-uncased-whole-word-masking-finetuned-squadz3bert-large-cased-whole-word-masking-finetuned-squadzbert-base-cased-finetuned-mrpczbert-base-german-dbmdz-casedzbert-base-german-dbmdz-uncasedzbert-base-finnish-cased-v1zbert-base-finnish-uncased-v1zbert-base-dutch-casedi do_lower_caseTFc CsRt}t|ddd}|}W5QRXt|D]\}}|d}|||<q2|S)z*Loads a vocabulary file into a dictionary.rutf-8encoding ) collections OrderedDictopen readlines enumeraterstrip)rvocabreadertokensindextokenrB/tmp/pip-unpacked-wheel-ymerj3tt/transformers/tokenization_bert.py load_vocabcs  rcCs|}|sgS|}|S)z@Runs basic whitespace cleaning and splitting on a piece of text.)stripsplit)textrrrrwhitespace_tokenizens r c seZdZdZeZeZeZ e Z d"fd d Z e d d Zd dZddZddZddZddZd#eeeeeeedddZd$eeeeeeeedddZd%eeeeeeedddZd d!ZZS)& BertTokenizeray Constructs a BERT tokenizer. Based on WordPiece. This tokenizer inherits from :class:`~transformers.PreTrainedTokenizer` which contains most of the methods. Users should refer to the superclass for more information regarding methods. Args: vocab_file (:obj:`string`): File containing the vocabulary. do_lower_case (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to lowercase the input when tokenizing. do_basic_tokenize (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to do basic tokenization before WordPiece. never_split (:obj:`bool`, `optional`, defaults to :obj:`True`): List of tokens which will never be split during tokenization. Only has an effect when :obj:`do_basic_tokenize=True` unk_token (:obj:`string`, `optional`, defaults to "[UNK]"): The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. sep_token (:obj:`string`, `optional`, defaults to "[SEP]"): The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for sequence classification or for a text and a question for question answering. It is also used as the last token of a sequence built with special tokens. pad_token (:obj:`string`, `optional`, defaults to "[PAD]"): The token used for padding, for example when batching sequences of different lengths. cls_token (:obj:`string`, `optional`, defaults to "[CLS]"): The classifier token which is used when doing sequence classification (classification of the whole sequence instead of per-token classification). It is the first token of the sequence when built with special tokens. mask_token (:obj:`string`, `optional`, defaults to "[MASK]"): The token used for masking values. This is the token used when training this model with masked language modeling. This is the token which the model will try to predict. tokenize_chinese_chars (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to tokenize Chinese characters. This should likely be deactivated for Japanese: see: https://github.com/huggingface/transformers/issues/328 TN[UNK][SEP][PAD][CLS][MASK]c  stjf||||| d| tj|s8td|t||_t dd|j D|_ ||_ |rxt||| d|_t|j|jd|_dS)N unk_token sep_token pad_token cls_token mask_tokenzCan't find a vocabulary file at path '{}'. To load the vocabulary from a Google pretrained model use `tokenizer = BertTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`cSsg|]\}}||fqSrr).0tokidsrrr sz*BertTokenizer.__init__..r never_splittokenize_chinese_chars)rr()super__init__ospathisfile ValueErrorformatrrrritems ids_to_tokensdo_basic_tokenizeBasicTokenizerbasic_tokenizerWordpieceTokenizerr(wordpiece_tokenizer) selfrr r=r2r(r)r*r+r,r3kwargs __class__rrr5s2  zBertTokenizer.__init__cCs t|jSN)lenrrBrrr vocab_sizeszBertTokenizer.vocab_sizecCst|jf|jSrF)dictrZadded_tokens_encoderrHrrr get_vocabszBertTokenizer.get_vocabcCsPg}|jr@|jj||jdD] }|j|D]}||q,qn |j|}|S)N)r2)r=r?tokenizeZall_special_tokensrAappend)rBr split_tokensrZ sub_tokenrrr _tokenizes zBertTokenizer._tokenizecCs|j||j|jS)z2 Converts a token (str) in an id using the vocab. )rgetr()rBrrrr_convert_token_to_idsz"BertTokenizer._convert_token_to_idcCs|j||jS)z=Converts an index (integer) in a token (str) using the vocab.)r<rPr()rBrrrr_convert_id_to_tokensz"BertTokenizer._convert_id_to_tokencCsd|dd}|S)z< Converts a sequence of tokens (string) in a single string.  z ##)joinreplacer)rBrZ out_stringrrrconvert_tokens_to_stringsz&BertTokenizer.convert_tokens_to_string) token_ids_0 token_ids_1returncCs@|dkr|jg||jgS|jg}|jg}|||||S)a Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A BERT sequence has the following format: - single sequence: ``[CLS] X [SEP]`` - pair of sequences: ``[CLS] A [SEP] B [SEP]`` Args: token_ids_0 (:obj:`List[int]`): List of IDs to which the special tokens will be added token_ids_1 (:obj:`List[int]`, `optional`, defaults to :obj:`None`): Optional second list of IDs for sequence pairs. Returns: :obj:`List[int]`: list of `input IDs <../glossary.html#input-ids>`__ with the appropriate special tokens. N cls_token_id sep_token_id)rBrXrYclsseprrr build_inputs_with_special_tokenss z.BertTokenizer.build_inputs_with_special_tokensF)rXrYalready_has_special_tokensrZcsz|r*|dk rtdttfdd|S|dk r`dgdgt|dgdgt|dgSdgdgt|dgS)a  Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding special tokens using the tokenizer ``prepare_for_model`` or ``encode_plus`` methods. Args: token_ids_0 (:obj:`List[int]`): List of ids. token_ids_1 (:obj:`List[int]`, `optional`, defaults to :obj:`None`): Optional second list of IDs for sequence pairs. already_has_special_tokens (:obj:`bool`, `optional`, defaults to :obj:`False`): Set to True if the token list is already formatted with special tokens for the model Returns: :obj:`List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token. Nz~You should not supply a second sequence if the provided sequence of ids is already formated with special tokens for the model.cs|jjfkrdSdS)Nrr)r]r\)xrHrrz7BertTokenizer.get_special_tokens_mask..rr)r9listmaprG)rBrXrYrarrHrget_special_tokens_masks.z%BertTokenizer.get_special_tokens_maskcCsV|jg}|jg}|dkr.t|||dgSt|||dgt||dgS)a Creates a mask from the two sequences passed to be used in a sequence-pair classification task. A BERT sequence pair mask has the following format: :: 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 | first sequence | second sequence | if token_ids_1 is None, only returns the first portion of the mask (0's). Args: token_ids_0 (:obj:`List[int]`): List of ids. token_ids_1 (:obj:`List[int]`, `optional`, defaults to :obj:`None`): Optional second list of IDs for sequence pairs. Returns: :obj:`List[int]`: List of `token type IDs <../glossary.html#token-type-ids>`_ according to the given sequence(s). Nrr)r]r\rG)rBrXrYr_r^rrr$create_token_type_ids_from_sequencess z2BertTokenizer.create_token_type_ids_from_sequencesc Csd}tj|r$tj|td}n|}t|dddX}t|jdddD]:\}}||krrt d ||}| |d |d 7}qNW5QRX|fS) a. Save the sentencepiece vocabulary (copy original file) and special tokens file to a directory. Args: vocab_path (:obj:`str`): The directory in which to save the vocabulary. Returns: :obj:`Tuple(str)`: Paths to the files saved. rrwr r cSs|dS)Nrr)kvrrrrcNrdz/BertTokenizer.save_vocabulary..)keyzsSaving vocabulary to {}: vocabulary indices are not consecutive. Please check that the vocabulary is not corrupted!rr) r6r7isdirrUVOCAB_FILES_NAMESrsortedrr;loggerwarningr:write)rBZ vocab_pathrrwriterrZ token_indexrrrsave_vocabulary=s  zBertTokenizer.save_vocabulary) TTNr"r#r$r%r&T)N)NF)N)__name__ __module__ __qualname____doc__rmvocab_files_namesPRETRAINED_VOCAB_FILES_MAPpretrained_vocab_files_mapPRETRAINED_INIT_CONFIGURATIONpretrained_init_configuration&PRETRAINED_POSITIONAL_EMBEDDINGS_SIZESmax_model_input_sizesr5propertyrIrKrOrQrRrWrintrr`boolrgrhrs __classcell__rrrDrr!wsV&%        r!c@sNeZdZdZdddZdddZdd Zdd d Zd d ZddZ ddZ dS)r>zDRuns basic tokenization (punctuation splitting, lower casing, etc.).TNcCs"|dkr g}||_||_||_dS)a Constructs a BasicTokenizer. Args: **do_lower_case**: Whether to lower case the input. **never_split**: (`optional`) list of str Kept for backward compatibility purposes. Now implemented directly at the base class level (see :func:`PreTrainedTokenizer.tokenize`) List of token not to split. **tokenize_chinese_chars**: (`optional`) boolean (default True) Whether to tokenize Chinese characters. This should likely be deactivated for Japanese: see: https://github.com/huggingface/pytorch-pretrained-BERT/issues/328 Nr1)rBr r2r3rrrr5]s zBasicTokenizer.__init__cCs|j|dk r|ng}||}|jr0||}t|}g}|D]6}|jrd||krd|}||}|| ||q@td |}|S)a Basic Tokenization of a piece of text. Split on "white spaces" only, for sub-word tokenization, see WordPieceTokenizer. Args: **never_split**: (`optional`) list of str Kept for backward compatibility purposes. Now implemented directly at the base class level (see :func:`PreTrainedTokenizer.tokenize`) List of token not to split. NrS) r2 _clean_textr3_tokenize_chinese_charsr r lower_run_strip_accentsextend_run_split_on_puncrU)rBrr2Z orig_tokensrNr output_tokensrrrrLqs    zBasicTokenizer.tokenizecCsBtd|}g}|D]"}t|}|dkr,q||qd|S)z$Strips accents from a piece of text.ZNFDZMnrT) unicodedata normalizecategoryrMrU)rBroutputcharcatrrrrs   z!BasicTokenizer._run_strip_accentscCs|dk r||kr|gSt|}d}d}g}|t|kr||}t|rX||gd}n |rf|gd}|d||d7}q*dd|DS) z&Splits punctuation on a piece of text.NrTFrcSsg|]}d|qS)rT)rU)r-rbrrrr0sz5BasicTokenizer._run_split_on_punc..)rerG_is_punctuationrM)rBrr2charsiZstart_new_wordrrrrrrs"    z!BasicTokenizer._run_split_on_punccCsTg}|D]@}t|}||r>|d|||dq||qd|S)z)Adds whitespace around any CJK character.rSrT)ord_is_chinese_charrMrUrBrrrcprrrrs     z&BasicTokenizer._tokenize_chinese_charscCs|dkr|dks|dkr |dks|dkr0|dks|dkr@|dks|d krP|d ks|d kr`|d ks|d krp|dks|dkr|dkrdSdS)z6Checks whether CP is the codepoint of a CJK character.iNii4iMiiߦii?i@ii iiiiiTFr)rBrrrrrsD  zBasicTokenizer._is_chinese_charcCsXg}|D]D}t|}|dks|dkst|r.qt|rB|dq||qd|S)zBPerforms invalid character removal and whitespace cleanup on text.rirSrT)r _is_control_is_whitespacerMrUrrrrrs  zBasicTokenizer._clean_text)TNT)N)N) rtrurvrwr5rLrrrrrrrrrr>Zs    r>c@s"eZdZdZdddZddZdS) r@zRuns WordPiece tokenization.dcCs||_||_||_dSrF)rr(max_input_chars_per_word)rBrr(rrrrr5szWordpieceTokenizer.__init__c Csg}t|D]}t|}t||jkr4||jq d}d}g}|t|krt|}d} ||krd|||} |dkrd| } | |jkr| } q|d8}qX| dkrd}q|| |}q@|r||jq ||q |S)aTokenizes a piece of text into its word pieces. This uses a greedy longest-match-first algorithm to perform tokenization using the given vocabulary. For example: input = "unaffable" output = ["un", "##aff", "##able"] Args: text: A single token or whitespace separated tokens. This should have already been passed through `BasicTokenizer`. Returns: A list of wordpiece tokens. FrNrT##rT) r rerGrrMr(rUrr) rBrrrrZis_badstartZ sub_tokensendZ cur_substrsubstrrrrrLs:       zWordpieceTokenizer.tokenizeN)r)rtrurvrwr5rLrrrrr@s r@cCs>|dks |dks |dks |dkr$dSt|}|dkr:dSdS)z1Checks whether `chars` is a whitespace character.rS r TZsF)rrrrrrrr s   rcCs8|dks|dks|dkrdSt|}|dr4dSdS)z.Checks whether `chars` is a control character.rrrFCT)rr startswithrrrrr,s   rcCsht|}|dkr|dksH|dkr(|dksH|dkr8|dksH|dkrL|dkrLd St|}|d rdd Sd S) z2Checks whether `chars` is a punctuation character.!/:@[`{~TPF)rrrr)rrrrrrr8s@  rc s<eZdZdZeZeZeZ e Z dfd d Z dd d Z ZS)BertTokenizerFastaY Constructs a "Fast" BERT tokenizer (backed by HuggingFace's `tokenizers` library). Bert tokenization is Based on WordPiece. This tokenizer inherits from :class:`~transformers.PreTrainedTokenizerFast` which contains most of the methods. Users should refer to the superclass for more information regarding methods. Args: vocab_file (:obj:`string`): File containing the vocabulary. do_lower_case (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to lowercase the input when tokenizing. unk_token (:obj:`string`, `optional`, defaults to "[UNK]"): The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. sep_token (:obj:`string`, `optional`, defaults to "[SEP]"): The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for sequence classification or for a text and a question for question answering. It is also used as the last token of a sequence built with special tokens. pad_token (:obj:`string`, `optional`, defaults to "[PAD]"): The token used for padding, for example when batching sequences of different lengths. cls_token (:obj:`string`, `optional`, defaults to "[CLS]"): The classifier token which is used when doing sequence classification (classification of the whole sequence instead of per-token classification). It is the first token of the sequence when built with special tokens. mask_token (:obj:`string`, `optional`, defaults to "[MASK]"): The token used for masking values. This is the token used when training this model with masked language modeling. This is the token which the model will try to predict. tokenize_chinese_chars (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to tokenize Chinese characters. This should likely be deactivated for Japanese: see: https://github.com/huggingface/transformers/issues/328 clean_text (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to clean the text before tokenization by removing any control characters and replacing all whitespaces by the classic one. tokenize_chinese_chars (:obj:`bool`, `optional`, defaults to :obj:`True`): Whether to tokenize Chinese characters. This should likely be deactivated for Japanese: see: https://github.com/huggingface/transformers/issues/328 Tr"r#r$r%r&rc s@tjt|||||| | || d f|||||d| ||_dS)N) rr(r)r+ clean_textZhandle_chinese_chars strip_accentsZ lowercasewordpieces_prefixr')r4r5rr ) rBrr r(r)r*r+r,rr3rrrCrDrrr5ws, zBertTokenizerFast.__init__NcCs,|jg||jg}|r(|||jg7}|SrFr[)rBrXrYrrrrr`sz2BertTokenizerFast.build_inputs_with_special_tokens) Tr"r#r$r%r&TTTr)N)rtrurvrwrmrxryrzr{r|r}r~r5r`rrrrDrrGs"*%r)rwrloggingr6rtypingrrZ tokenizersrZtokenization_utilsrr getLoggerrtrormryr}r{rr r!objectr>r@rrrrrrrrs    d <