U 9%eN@sdZddlZddlZddlZddlZddlmZmZmZm Z ddl m Z ddl m Z e eZddd d Zd d id d id did Zd diZd ddgddddiZddZddZddZGddde ZdS)zTokenization classes for FSMT.N)DictListOptionalTuple)PreTrainedTokenizer)loggingzvocab-src.jsonzvocab-tgt.jsonz merges.txt)src_vocab_filetgt_vocab_file merges_filezstas/tiny-wmt19-en-dezHhttps://huggingface.co/stas/tiny-wmt19-en-de/resolve/main/vocab-src.jsonzHhttps://huggingface.co/stas/tiny-wmt19-en-de/resolve/main/vocab-tgt.jsonzDhttps://huggingface.co/stas/tiny-wmt19-en-de/resolve/main/merges.txtiende)langsZmodel_max_lengthZspecial_tokens_map_fileZfull_tokenizer_filecCs6t}|d}|ddD]}|||f|}q|S)z Return set of symbol pairs in a word. word is represented as tuple of symbols (symbols being variable-length strings) rN)setadd)wordpairsZ prev_charcharri/var/www/html/Darija-Ai-API/env/lib/python3.8/site-packages/transformers/models/fsmt/tokenization_fsmt.py get_pairs9s rcCs|dd}tdd|}|dd}|dd}|dd}|d d }|d d }|d d }|dd}|dd}|dd}|dd}|dd}|dd}|dd}|dd}|dd}|dd}|dd}|d d!}|d"d#}|d$d%}|d&d'}|d(d)}|d*d+}|d,d-}td.d|}|d/d0}|d1d2}|d3d4}|d5d6}|d7d8}|d9d:}|d;d<}|d=d>}|d?d@}|S)Azz Port of https://github.com/moses-smt/mosesdecoder/blob/master/scripts/tokenizer/replace-unicode-punctuation.perl u，,u。\s*z. u、u”"u“u∶:u：u？?u《u》u）)u！!u（(u；;u１1u」u「u０0u３3u２2u５5u６6u９9u７7u８8u４4u．\s*u～~u’'u…z...u━-u〈u【[u】]u％%)replaceresub)textrrrreplace_unicode_punctFsJ                                  r6cCs8g}|D]$}t|}|dr"q||qd|S)zw Port of https://github.com/moses-smt/mosesdecoder/blob/master/scripts/tokenizer/remove-non-printing-char.perl C) unicodedatacategory startswithappendjoin)r5outputrcatrrrremove_non_printing_charqs   r@c sreZdZdZeZeZeZ e Z ddgZ d;fd d Z eeefd d dZeed ddZddZddZddZddZeddZeddZddZdd Zd!d"Zde!ee"e!ee$e!ed/fd0d1 Z%d?e!ee"e!ee!ed,d2d3Z&d@ee"ee'ed4d5d6Z(d7d8Z)d9d:Z*Z+S)A FSMTTokenizera Construct an FAIRSEQ Transformer tokenizer. Based on Byte-Pair Encoding. The tokenization process is the following: - Moses preprocessing and tokenization. - Normalizing all inputs text. - The arguments `special_tokens` and the function `set_special_tokens`, can be used to add additional symbols (like "__classify__") to a vocabulary. - The argument `langs` defines a pair of languages. This tokenizer inherits from [`PreTrainedTokenizer`] which contains most of the main methods. Users should refer to this superclass for more information regarding those methods. Args: langs (`List[str]`): A list of two languages to translate from and to, for instance `["en", "ru"]`. src_vocab_file (`str`): File containing the vocabulary for the source language. tgt_vocab_file (`st`): File containing the vocabulary for the target language. merges_file (`str`): File containing the merges. do_lower_case (`bool`, *optional*, defaults to `False`): Whether or not to lowercase the input when tokenizing. unk_token (`str`, *optional*, defaults to `""`): 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. bos_token (`str`, *optional*, defaults to `""`): The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token. When building a sequence using special tokens, this is not the token that is used for the beginning of sequence. The token used is the `cls_token`. sep_token (`str`, *optional*, defaults to `""`): 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 (`str`, *optional*, defaults to `""`): The token used for padding, for example when batching sequences of different lengths. Z input_idsZattention_maskNFc  sfz ddl} Wntk r(tdYnX| |_||_||_||_||_i|_i|_i|_ |rxt |dkrx|\|_ |_ nt d|dt|dd} t| |_W5QRXt|dd$} t| }dd |D|_W5QRXt|dd}|d dd }W5QRXd d |D}tt|tt ||_i|_tjf||||||||| d | dS)NrnYou need to install sacremoses to use XLMTokenizer. See https://pypi.org/project/sacremoses/ for installation.zFarg `langs` needs to be a list of 2 langs, e.g. ['en', 'ru'], but got z. Usually that means that tokenizer can't find a mapping for the given model path in PRETRAINED_VOCAB_FILES_MAP, and other maps of this tokenizer.utf-8encodingcSsi|]\}}||qSrr.0kvrrr sz*FSMTTokenizer.__init__.. cSs g|]}t|ddqS)NrG)tuplesplit)rLmergerrr sz*FSMTTokenizer.__init__..) rr r r do_lower_case unk_token bos_token sep_token pad_token) sacremoses ImportErrorsmr r r rVcache_moses_punct_normalizercache_moses_tokenizercache_moses_detokenizerlensrc_langtgt_lang ValueErroropenjsonloadencoderitemsdecoderreadrSdictziprange bpe_rankscachesuper__init__)selfrr r r rVrWrXrYrZkwargsr[Zsrc_vocab_handleZtgt_vocab_handle tgt_vocabZ merges_handleZmerges __class__rrrrsT       zFSMTTokenizer.__init__)returncCs|SN) get_src_vocabrsrrr get_vocabszFSMTTokenizer.get_vocabcCs|jSry)src_vocab_sizer{rrr vocab_sizeszFSMTTokenizer.vocab_sizecCs2||jkr"|jj|d}||j|<|j||SNlang)r^r]ZMosesPunctNormalizer normalize)rsr5rZpunct_normalizerrrrmoses_punct_norms  zFSMTTokenizer.moses_punct_normcCs:||jkr"|jj|d}||j|<|j|j|ddddS)NrTF)Zaggressive_dash_splitsZ return_strescape)r_r]ZMosesTokenizertokenize)rsr5rZmoses_tokenizerrrrmoses_tokenize s   zFSMTTokenizer.moses_tokenizecCs2||jkr"|jj|d}||j|<|j||Sr)r`r]ZMosesDetokenizerZ detokenize)rstokensrZmoses_detokenizerrrrmoses_detokenizes  zFSMTTokenizer.moses_detokenizecCs t|}|||}t|}|Sry)r6rr@)rsr5rrrrmoses_pipelines zFSMTTokenizer.moses_pipelinecCs t|jSry)rarhr{rrrr} szFSMTTokenizer.src_vocab_sizecCs t|jSry)rarjr{rrrtgt_vocab_size$szFSMTTokenizer.tgt_vocab_sizecCst|jf|jSry)rlrhZadded_tokens_encoderr{rrrrz(szFSMTTokenizer.get_src_vocabcCst|jf|jSry)rlrjZadded_tokens_decoderr{rrr get_tgt_vocab+szFSMTTokenizer.get_tgt_vocabc st|dd|ddf}|jkr2j|St|}|sF|dSt|fddd}|jkrhqf|\}}g}d}|t|krcsj|tdS)Ninf)rogetfloat)pairr{rr8z#FSMTTokenizer.bpe..keyrrrG z z ) rRrprminroraindexrdextendr<r=) rstokenrrZbigramfirstsecondZnew_wordijrr{rbpe.sF    2      zFSMTTokenizer.bper cCsn|j}|jr|}|r"|}n|j||d}|j||d}g}|D]"}|rF|t||dqF|S)av Tokenize a string given language code using Moses. Details of tokenization: - [sacremoses](https://github.com/alvations/sacremoses): port of Moses - Install with `pip install sacremoses` Args: - lang: ISO language code (default = 'en') (string). Languages should belong of the model supported languages. However, we don't enforce it. - bypass_tokenizer: Allow users to preprocess and tokenize the sentences externally (default = False) (bool). If True, we only apply BPE. Returns: List of tokens. rr) rbrVlowerrSrrrlistr)rsr5rZbypass_tokenizerZ split_tokensrrrr _tokenizeZs zFSMTTokenizer._tokenizecCs|j||j|jS)z0Converts a token (str) in an id using the vocab.)rhrrW)rsrrrr_convert_token_to_idsz"FSMTTokenizer._convert_token_to_idcCs|j||jS)z=Converts an index (integer) in a token (str) using the vocab.)rjrrW)rsrrrr_convert_id_to_tokensz"FSMTTokenizer._convert_id_to_tokencCs.dd|D}d|}|||j}|S)z:Converts a sequence of tokens (string) in a single string.cSs g|]}|ddddqS)rr8r)r2)rLtrrrrUsz:FSMTTokenizer.convert_tokens_to_string..r8)r=rSrrc)rsrr5rrrconvert_tokens_to_stringsz&FSMTTokenizer.convert_tokens_to_string) token_ids_0 token_ids_1rxcCs(|jg}|dkr||S||||S)a Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A FAIRSEQ Transformer sequence has the following format: - single sequence: ` X ` - pair of sequences: ` A B ` Args: token_ids_0 (`List[int]`): List of IDs to which the special tokens will be added. token_ids_1 (`List[int]`, *optional*): Optional second list of IDs for sequence pairs. Returns: `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens. N) sep_token_idrsrrseprrr build_inputs_with_special_tokenssz.FSMTTokenizer.build_inputs_with_special_tokens)rralready_has_special_tokensrxcsZ|rtj||ddS|dk rFdgt|dgdgt|dgSdgt|dgS)a Retrieve 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` method. Args: token_ids_0 (`List[int]`): List of IDs. token_ids_1 (`List[int]`, *optional*): Optional second list of IDs for sequence pairs. already_has_special_tokens (`bool`, *optional*, defaults to `False`): Whether or not the token list is already formatted with special tokens for the model. Returns: `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token. T)rrrNrr)rqget_special_tokens_maskra)rsrrrrvrrrs(z%FSMTTokenizer.get_special_tokens_maskcCsF|jg}|dkr"t||dgSt||dgt||dgS)a Create a mask from the two sequences passed to be used in a sequence-pair classification task. A FAIRSEQ Transformer 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`, this method only returns the first portion of the mask (0s). Args: token_ids_0 (`List[int]`): List of IDs. token_ids_1 (`List[int]`, *optional*): Optional second list of IDs for sequence pairs. Returns: `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s). Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An FAIRSEQ_TRANSFORMER sequence pair mask has the following format: Nrr)rrarrrr$create_token_type_ids_from_sequencessz2FSMTTokenizer.create_token_type_ids_from_sequences)save_directoryfilename_prefixrxc Cstj|s"td|ddStj||r6|dndtd}tj||rX|dndtd}tj||rz|dndtd}t|dd d $}|t j |j d d d ddW5QRXt|dd d 6}dd|j D}|t j |d d d ddW5QRXd}t|dd d d} t|j dddD]F\} } || kr\td|d| }| d| d|d7}q4W5QRX|||fS)NzVocabulary path (z) should be a directoryr,r8r r r wrHrIrGTF)indent sort_keys ensure_asciirPcSsi|]\}}||qSrrrKrrrrOsz1FSMTTokenizer.save_vocabulary..rcSs|dS)Nrr)kvrrrrrz/FSMTTokenizer.save_vocabulary..rzSaving vocabulary to zZ: BPE merge indices are not consecutive. Please check that the tokenizer is not corrupted!rr)ospathisdirloggererrorr=VOCAB_FILES_NAMESrewriterfdumpsrhrjrisortedrowarning) rsrrr r r frurwriterZ bpe_tokensZ token_indexrrrsave_vocabularys> (&  zFSMTTokenizer.save_vocabularycCs|j}d|d<|S)Nr])__dict__copy)rsstaterrr __getstate__s zFSMTTokenizer.__getstate__cCs:||_z ddl}Wntk r.tdYnX||_dS)NrrF)rr[r\r])rsdr[rrr __setstate__s  zFSMTTokenizer.__setstate__) NNNNFrBrCrDrE)r F)N)NF)N)N),__name__ __module__ __qualname____doc__rZvocab_files_namesPRETRAINED_VOCAB_FILES_MAPZpretrained_vocab_files_mapPRETRAINED_INIT_CONFIGURATIONZpretrained_init_configuration&PRETRAINED_POSITIONAL_EMBEDDINGS_SIZESZmax_model_input_sizesZmodel_input_namesrrrstrintr|propertyr~rrrrr}rrzrrrrrrrrrboolrrrrrr __classcell__rrrvrrAsr-C  , '      !$rA)rrfrr3r9typingrrrrZtokenization_utilsrutilsrZ get_loggerrrrrrrrr6r@rArrrrs@      +