]hoN4ddlmZddlZddlZddlZddlmZddlmZm Z m Z ddl m Z m Z  ddlmZddlZddlmZmZmZmZmZmZddlmZdd lmZdd lmZej>e Z!er erdd l"m#Z#dd Z$Gd deZ%y#e$r ddlmZYdwxYw)) annotationsN)Path) TYPE_CHECKINGAnyCallable)load_onnx_modelload_openvino_model)Self) AutoConfig AutoModel AutoTokenizer MT5ConfigPretrainedConfigT5Config)is_peft_available)find_adapter_config_file) InputModule PeftConfigcdfd }|S)Ncttjt|z dt|z fi|S)NT)exist_ok)osmakedirsr)save_directorykwargs_save_pretrained_fn subfolders f/var/www/html/eduruby.in/venv/lib/python3.12/site-packages/sentence_transformers/models/Transformer.pywrapperz)_save_pretrained_wrapper..wrappers5 D(94tD"4#7)#CNvNN)rz str | PathreturnNone)rrr s`` r_save_pretrained_wrapperr%sO Nr!ceZdZUdZdZded<ddgZded<d Zd ed < d dfd Z dd Z ddZ ddZ ddZ ddZ ddZd dZ d! d"dZd!d#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__r,backend _load_config _load_modellistinspect signature auto_modelforward parameterssetmodel_forward_paramsr from_pretrained tokenizerhasattrr7minr8r5r+ __class____name__tokenizer_class)selfmodel_name_or_pathr+ model_argstokenizer_args config_argsr6r,tokenizer_name_or_pathr;r7 is_peft_modelrErJs rr:zTransformer.__init__As *  J  !N  K $ 1 12DiQXZe f +VYeZde $G$5$5doo6M6M$N$Y$YZ$'(<$=A % !  %*?!$T__%;%;%S%SUYUcUcUtUt!u, ! -59^^5M5M5V5VDOO " " 2 .r!c @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)r6rUrVrWzgLoading 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.rrr6T) rgetr Exception ValueErrorpeftrrFr )rMrNr6r;rQrs rr<zTransformer._load_configzs %"#!oog.$4!,1CU!K   %&}'! w (-:--.@eKe[degkk k))*<a aW`achhhr!c |dk(r|rdD]}|j|dt|tr|j|||fi|yt|tr|j |||fi|yt j|f||d||_y|dk(rtd ||dd||_y|dk(rtd ||dd||_ytd |d ) aLoads the transformers or PEFT model into the `auto_model` attribute Args: model_name_or_path (str): The model name on Hugging Face (e.g. 'sentence-transformers/all-MiniLM-L6-v2') or the path to a local model directory. config ("PeftConfig" | PretrainedConfig): The model configuration. cache_dir (str | None): The cache directory to store the model configuration. backend (str): The backend used for model inference. Can be `torch`, `onnx`, or `openvino`. is_peft_model (bool): Whether the model is a PEFT model. model_args (dict[str, Any]): Keyword arguments passed to the Hugging Face Transformers model. rX)rVNr7r6onnxzfeature-extraction)rNr7 task_nameopenvinozUnsupported backend 'z6'. `backend` should be `torch`, `onnx`, or `openvino`.r$) pop isinstancer_load_t5_modelr_load_mt5_modelr rFrArr r[)rMrNr7r6r;rSrOadapter_only_kwargs rr=zTransformer._load_models( g *6=&NN#5t<=&(+###$6 XZXFI.$$$%7YjY"+";";&#/5#NX# -#5. DO  "1#5. DO4WI=stu ur!c Vddlm}dg|_|j|f||d||_y)Loads the encoder model from T5r)T5EncoderModel decoder.*r^N) transformersri"_keys_to_ignore_on_load_unexpectedrFrA)rMrNr7r6rOris rrdzTransformer._load_t5_models8/=HM98.88  '- FP r!c Vddlm}dg|_|j|f||d||_y)rhr)MT5EncoderModelrjr^N)rkrnrlrFrA)rMrNr7r6rOrns rrezTransformer._load_mt5_models80>I]:9/99  '- FP r!c|dt|j|jjjdS)Nz Transformer() architecture))dictget_config_dictrArJrKrMs r__repr__zTransformer.__repr__s3d4#7#7#9HaHaHjHjkllmnnr!c |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)PeftModelForFeatureExtractionr3)device)dim hidden_statesall_layer_embeddingsr$)itemsrErArr\ryrcactive_peft_configis_prompt_learningsizerXonesnum_virtual_tokensrzcatr7output_hidden_states) rMfeaturesrkeyvaluetrans_featuresoutputsrxry batch_sizer3prefix_attention_masks rrBzTransformer.forwards2,8@~~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;cB|jjjS)N)rAr7 hidden_sizerts rget_word_embedding_dimensionz(Transformer.get_word_embedding_dimensions%%111r!c 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) rcr)rrnextiterrappendstripr,lowerupdaterGr+) rMtextsroutput to_tokenizelookuptext_keytextbatch1batch2 text_tuplecolss rtokenizezTransformer.tokenize!sr 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)rAsave_pretrainedrG save_config)rM output_pathrrs rsavezTransformer.saveIs: '' HZ'[ &&{3 %r!c N|j||||||||| | |  } |dd|i| S)N) rNrrU cache_folderrVrWtrust_remote_code model_kwargstokenizer_kwargs config_kwargsr;rNr$)_load_init_kwargs)clsrNrrUrrVrWrrrrr;r init_kwargss rloadzTransformer.loadNsP$++1%-/%-',  H&8HKHHr!c  |j||||||} |||||d}d| vri| d<d| vri| d<d| vri| d<| dj|| dj|| dj||r| dj|| r| dj| | r| dj| i| || dS)N)rNrrUrrVrW)rrUrVrWrrOrPrQ)r6r;) load_configr)rrNrrUrrVrWrrrrr;rr7 hub_kwargss rrzTransformer._load_init_kwargsos$1%- ! # 0!2   v %#%F< 6 )')F# $  &$&F= ! |##J/ '' 3}$$Z0  < ' ' 5  # $ + +,< =  = ! ( ( 7H&H|HHr!c |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)rNrconfig_filenamerUrrVrWrOrrPrQ)r9rrb) rrNrrrUrrVrWconfig_filenamesr7rJs rrzTransformer.load_configs   0 OW(#5# /)!!1)F  6 !&9VL=Q&Q < $ $%8 9 v %*=HXAY*Y # $ ( ()< = F "':f]>S'S = ! % %&9 : r!)NNNNNFNrX)rNr)r+z int | NonerOdict[str, Any] | NonerPrrQrr6 str | Noner,r.rRrr;r)r"r#) rNr)r6rr;r)rQdict[str, Any]r"z*tuple[PeftConfig | PretrainedConfig, bool]) rNr)r7zPeftConfig | PretrainedConfigr6r)r;r)rSr.r"r#)rNr)r7rr6r)r"r#)r"r))rdict[str, torch.Tensor]r"r)r"int)T)rz.list[str] | list[dict] | list[tuple[str, str]]rz str | boolr"r)rr)rr.r"r#) NNNFFNNNrX)rNr)rr)rUbool | str | NonerrrVrrWr.rr.rrrrrrr;r)r"r )rNr)rr)rUrrrrVrrWr.rr.rrrrrrr;r)r"r)rNNNNF)rNr)rr)rrrUrrrrVrrWr.r"r)rK __module__ __qualname____doc__r*__annotations__r-r/r:r<r=rdrerurBrrr classmethodrrr __classcell__)rJs@rr'r'%sc.8c7.@K@L$ &*,004-1 $#-17W7W#7W* 7W . 7W + 7W7W7W!+7W7W 7Wr(i"%(i2<(iGJ(iYg(i 3(iT2v2v.2v 2v  2v  2v 2vh  o/b2\`&C&NX& &P&  #'#'#!&"'.226/3II I ! I ! III I,I0I-II !II@ #'#'#!&"'.226/37I7I 7I ! 7I ! 7I7I7I 7I,7I07I-7I7I !7I7Ir&*#'#'#!&+++$ + ! + ! +++ ++r!r')rrrr)r"zCallable[..., None])& __future__rr?loggingrpathlibrtypingrrrsentence_transformers.backendrr r ImportErrortyping_extensionsrXrkr r r rrrtransformers.utils.import_utilsrtransformers.utils.peft_utilsr(sentence_transformers.models.InputModuler getLoggerrKloggerr\rr%r'r$r!rrs~" //N' dd=B@   8 $&(p+p1'&'sB BB