LioXddlmZddlZddlZddlZddlmZddlmZddl m Z ddl m Z m Z ddlmZmZ ddl mZddlZdd lmZmZmZmZmZmZdd lmZdd lmZdd l m!Z!e r erdd l"m#Z# ddlm$Z$m%Z% ddlm&Z&ejNe(Z)edZ*ddZ+Gdde!Z,y#e$r ddlmZY|wxYw#e$rGddZ$GddZ%Y]wxYw#e$r GddZ&YkwxYw)) annotationsN)Callable)contextmanager)Path) TYPE_CHECKINGAny)load_onnx_modelload_openvino_model)Self) AutoConfig AutoModel AutoTokenizer MT5ConfigPretrainedConfigT5Config)is_peft_available)find_adapter_config_file) InputModule PeftConfig)T5Gemma2ConfigT5Gemma2TextConfigc eZdZy)rN__name__ __module__ __qualname__n/var/www/html/lcp-python-backend/venv/lib/python3.12/site-packages/sentence_transformers/models/Transformer.pyrr  rrc eZdZy)rNrrrr rr#r!rr) T5GemmaConfigc eZdZy)r#Nrrrr r#r#+r!rr#c +>K|Dcic]}|t||d}} |jD]\}}t|||d|jD]\}}t|||ycc}w#|jD]\}}t|||wxYww)N)getattritemssetattr)cls overridesname originalsvalues r set_temporary_class_attrsr.2s.wrapper?s5 D(94tD"4#7)#CNvNNr)r4z str | PathreturnNoner)r6r7r8s`` r _save_pretrained_wrapperr;>sO NrceZdZUdZdZded<ddgZded<d Zd ed < d dfd Z dd Z ddZ ddZ ddZ ddZ d ddZdd dZe d! d"dZe d! d#dZe d$ d%fd ZxZS)& TransformeraHugging Face AutoModel to generate token embeddings. Loads the correct class, e.g. BERT / RoBERTa etc. Args: model_name_or_path: Hugging Face models name (https://huggingface.co/models) max_seq_length: Truncate any inputs longer than max_seq_length model_args: Keyword arguments passed to the Hugging Face Transformers model tokenizer_args: Keyword arguments passed to the Hugging Face Transformers tokenizer config_args: Keyword arguments passed to the Hugging Face Transformers config cache_dir: Cache dir for Hugging Face Transformers to store/load models do_lower_case: If true, lowercases the input (independent if the model is cased or not) tokenizer_name_or_path: Name or path of the tokenizer. When None, then model_name_or_path is used backend: Backend used for model inference. Can be `torch`, `onnx`, or `openvino`. Default is `torch`. sentence_bert_config.jsonstrconfig_file_namemax_seq_length do_lower_casez list[str] config_keysTbool save_in_rootc Jt |||_| |_|i}|i}|i}|j ||| |\} } |j || || | fi|t tj|jjj} t| hdz|_ | d|vr||d<tj||n|fd|i||_|t#|jdrtt#|jj$drTt#|j dr>t'|jj$j(|j j*}||_|:|j j.j0|jj$_yy)N> input_ids inputs_embedsattention_masktoken_type_idsmodel_max_length cache_dirconfigmax_position_embeddings)super__init__rBbackend _load_config _load_modellistinspect signature auto_modelforward parameterssetmodel_forward_paramsrfrom_pretrained tokenizerhasattrrMminrNrKrA __class__rtokenizer_class)selfmodel_name_or_pathrA model_argstokenizer_args config_argsrLrBtokenizer_name_or_pathrQrM is_peft_modelr[r`s r rPzTransformer.__init__bs *  J  !N  K $ 1 12DiQXZe f +VYeZde $G$5$5doo6M6M$N$Y$YZ$'(<$=A % !  %*?!$T__%;%;%S%SUYUcUcUtUt!u, ! -59^^5M5M5V5VDOO " " 2 .rc @t|||jd|jd|jdd Dts td|dk7r t dd d lm}|j|fi|d |id fStj|fi|d |idfS) aLoads the transformers or PEFT configuration Args: model_name_or_path (str): The model name on Hugging Face (e.g. 'sentence-transformers/all-MiniLM-L6-v2') or the path to a local model directory. cache_dir (str | None): The cache directory to store the model configuration. backend (str): The backend used for model inference. Can be `torch`, `onnx`, or `openvino`. config_args (dict[str, Any]): Keyword arguments passed to the Hugging Face Transformers config. Returns: tuple[PretrainedConfig, bool]: The model configuration and a boolean indicating whether the model is a PEFT model. tokenrevisionlocal_files_onlyF)rLrjrkrlzgLoading a PEFT model requires installing the `peft` package. You can install it via `pip install peft`.torchaPEFT models can currently only be loaded with the `torch` backend. To use other backends, load the model with `backend="torch"`, call `model.transformers_model.merge_and_unload()`, save that model with `model.save_pretrained()` and then load the model with the desired backend.rrrLT) rgetr Exception ValueErrorpeftrr\r )rbrcrLrQrfrs r rRzTransformer._load_configs %"#!oog.$4!,1CU!K   %&}'! w (-:--.@eKe[degkk k))*<a aW`achhhrc 8|dk(r|rdD]}|j|dt|tr9ddlm}t |dg5|j |f||d||_dddyt|tr9dd lm } t | dg5| j |f||d||_dddyt|tr@dd lm } d |_ t | dg5| j |f||d||_dddyt|trDdd lm} t | d dg5| j |f|j |d||_dddyt|t"r"dd lm} | j |f||d||_yt%j |f||d||_y|dk(rt'd||dd||_y|dk(rt)d||dd||_yt+d|d#1swYyxYw#1swYyxYw#1swYyxYw#1swYyxYw)aLoads the transformers or PEFT model into the `auto_model` attribute Args: model_name_or_path (str): The model name on Hugging Face (e.g. 'sentence-transformers/all-MiniLM-L6-v2') or the path to a local model directory. config ("PeftConfig" | PretrainedConfig): The model configuration. cache_dir (str | None): The cache directory to store the model configuration. backend (str): The backend used for model inference. Can be `torch`, `onnx`, or `openvino`. is_peft_model (bool): Whether the model is a PEFT model. model_args (dict[str, Any]): Keyword arguments passed to the Hugging Face Transformers model. rm)rkNr)T5EncoderModelz decoder.*)"_keys_to_ignore_on_load_unexpected)rMrL)MT5EncoderModel)T5GemmaEncoderModelF)T5Gemma2Encoderz model.encoder)base_model_prefixrtonnxzfeature-extraction)rcrM task_nameopenvinozUnsupported backend 'z6'. `backend` should be `torch`, `onnx`, or `openvino`.r)pop isinstancer transformersrsr.r\rWrrur#rvis_encoder_decoderr.transformers.models.t5gemma2.modeling_t5gemma2rwencoderrr r r rp) rbrcrMrLrQrhrdadapter_only_kwargrsrurvrws r rSzTransformer._load_models( g *6=&NN#5t<=&(+7.~cnbop&Dn&D&D*'39Y'R\'DOFI.8.docpq&Eo&E&E*'39Y'R\'DOFM2<,1)./Bhsgtu&I&9&I&I*'39Y'R\'DOFN3Z.#&58C} 'Fo&E&E*'39>>Y'Zd'DO F$67Z"A/"A"A&#/5#NX##,";";&#/5#NX# -#5. DO  "1#5. DO4WI=stu uys0G,G8H)&H,G58HH Hc|dt|j|jjjdS)Nz Transformer() architecture))dictget_config_dictrWr`r)rbs r __repr__zTransformer.__repr__"s3d4#7#7#9HaHaHjHjkllmnnrc |jDcic]\}}||jvs||}}}|jd i||ddi}|d}||d<trddlm}t |j|r|jjjrr|jd} |d} tj| |jjj| j} tj| | fd |d<|jjj r d |vr|d |d <|Scc}}w) a^ Forward pass through the transformer model. This method processes the input features through the underlying transformers model and returns the token embeddings along with any other relevant outputs. Notes: - Only passes arguments that are expected by the underlying transformer model Args: features (dict[str, torch.Tensor]): Input features dictionary containing at least 'input_ids' and 'attention_mask'. May also contain other tensors required by the underlying transformer model. **kwargs: Additional keyword arguments to pass to the underlying transformer model. Returns: dict[str, torch.Tensor]: Updated features dictionary containing the input features, plus: - 'token_embeddings': Token-level embeddings from the transformer model - 'attention_mask': Possibly modified attention mask if using PeftModel with prompt learning - 'all_layer_embeddings': If the model outputs hidden states, contains embeddings from all layers return_dictTrtoken_embeddings)PeftModelForFeatureExtractionrI)device)dim hidden_statesall_layer_embeddingsr)r'r[rWrrqrr}active_peft_configis_prompt_learningsizermonesnum_virtual_tokensrcatrMoutput_hidden_states) rbfeaturesr5keyr-trans_featuresoutputsrr batch_sizerIprefix_attention_masks r rXzTransformer.forward%s2,8@~~7Gle3RVRkRkKk#u*ll!$//ONOfO$O"1:'7#$   :4??,IJOO66II-2215 !)*:!;(-  B B U U^l^s^s)%.3YY8M~7^de-f)* ?? ! ! 6 6?g;U/6/GH+ ,3ms D;D;c |jjj}|jS#t$r$|jj}Y|jSwxYw)aGet the output embedding dimension from the transformer model. Returns: int: The hidden dimension size of the model's embeddings. Raises: AttributeError: If the embedding dimension cannot be determined from the model config. )rWrMget_text_configAttributeError hidden_size)rb text_configs r get_word_embedding_dimensionz(Transformer.get_word_embedding_dimensionVsY 1//00@@BK&&& 1//00K&&& 1s$2AAc i}t|dtr|g}nt|dtr\g}g|d<|D]L}tt |j \}}|j ||dj |N|g}n7gg} }|D]*} |j | d| j | d,|| g}|D cgc])} | D cgc]} t| jc} +}} } |jr-|D cgc] } | D cgc]} | jc} "}} } |j|j||dd|jd|Scc} wcc} } wcc} wcc} } w)z-Tokenizes a text and maps tokens to token-idsr text_keysr longest_firstpt)padding truncationreturn_tensors max_length) r}r?rnextiterr'appendstriprBlowerupdater]rA) rbtextsroutput to_tokenizelookuptext_keytextbatch1batch2 text_tuplecolss r tokenizezTransformer.tokenizehsr eAh $ 'K a$ 'K"$F;  5!%d6<<>&:!;$""4({#**84 5'-KFF# -  jm, jm, -"6*KALL41A 4L L   ?JKs3!AGGI3KKK DNN*#..    5L4Ks0 E$ E .E$ E/E* 'E/E$*E/c |jj|||jj||j|y)N)safe_serialization)rWsave_pretrainedr] save_config)rb output_pathrr5s r savezTransformer.saves: '' HZ'[ &&{3 %rc N|j||||||||| | |  } |dd|i| S)N) rcr7rj cache_folderrkrltrust_remote_code model_kwargstokenizer_kwargs config_kwargsrQrcr)_load_init_kwargs)r)rcr7rjrrkrlrrrrrQr5 init_kwargss r loadzTransformer.loadsP$++1%-/%-',  H&8HKHHrc  |j||||||} |||||d}d| vri| d<d| vri| d<d| vri| d<| dj|| dj|| dj||r| dj|| r| dj| | r| dj| i| || dS)N)rcr7rjrrkrl)r7rjrkrlrrdrerf)rLrQ) load_configr)r)rcr7rjrrkrlrrrrrQr5rM hub_kwargss r rzTransformer._load_init_kwargss$1%- ! # 0!2   v %#%F< 6 )')F# $  &$&F= ! |##J/ '' 3}$$Z0  < ' ' 5  # $ + +,< =  = ! ( ( 7H&H|HHrc |r|gngd}|D]}t ||||||||} | snd vrd| dvr| djdd| vrd| dvr| djdd| vrd| dvr| djd| S)N)r>zsentence_roberta_config.jsonzsentence_distilbert_config.jsonzsentence_camembert_config.jsonzsentence_albert_config.jsonz sentence_xlm-roberta_config.jsonzsentence_xlnet_config.json)rcr7config_filenamerjrrkrlrdrrerf)rOrr|) r)rcr7rrjrrkrlconfig_filenamesrMr`s r rzTransformer.load_configs   0 OW(#5# /)!!1)F  6 !&9VL=Q&Q < $ $%8 9 v %*=HXAY*Y # $ ( ()< = F "':f]>S'S = ! % %&9 : r)NNNNNFNrm)rcr?rAz int | Nonerddict[str, Any] | NonererrfrrL str | NonerBrDrgrrQr?r9r:) rcr?rLrrQr?rfdict[str, Any]r9z*tuple[PeftConfig | PretrainedConfig, bool]) rcr?rMzPeftConfig | PretrainedConfigrLr?rQr?rhrDr9r:)r9r?)rdict[str, torch.Tensor]r9r)r9int)T)rz.list[str] | list[dict] | list[tuple[str, str]]rz str | boolr9r)rr?rrDr9r:) NNNFFNNNrm)rcr?r7r?rjbool | str | NonerrrkrrlrDrrDrrrrrrrQr?r9r )rcr?r7r?rjrrrrkrrlrDrrDrrrrrrrQr?r9r)rNNNNF)rcr?r7r?rrrjrrrrkrrlrDr9r)rrr__doc__r@__annotations__rCrErPrRrSrrXrrr classmethodrrr __classcell__)r`s@r r=r=Fs`.8c7.@K@L$ &*,004-1 $#-17W7W#7W* 7W . 7W + 7W7W7W!+7W7W 7Wr(i"%(i2<(iGJ(iYg(i 3(iT[v[v.[v [v  [v  [v [vzo/b'&\`&C&NX& &P&  #'#'#!&"'.226/3II I ! I ! III I,I0I-II !II@ #'#'#!&"'.226/37I7I 7I ! 7I ! 7I7I7I 7I,7I07I-7I7I !7I7Ir&*#'#'#!&+++$ + ! + ! +++ ++rr=)r6rr7r?r9zCallable[..., None])- __future__rrUloggingr2collections.abcr contextlibrpathlibrtypingrrsentence_transformers.backendr r r ImportErrortyping_extensionsrmr~r r rrrrtransformers.utils.import_utilsrtransformers.utils.peft_utilsr(sentence_transformers.models.InputModulerrqrrrr# getLoggerrloggerr.r;r=rrr rs" $%%N' dd=B@&( ? *   8 $&&V+Vo'&'          s5B+2B<;C+ B98B9<CCC-,C-