hoNLSSKJr SSKrSSKrSSKrSSKJr SSKJrJ r J r SSK J r J r SSKJr SSKrSSKJrJrJrJrJrJr SSKJr SS KJr SS KJr \R>"\ 5r!\(a\"5(aSS K"J#r# SS jr$"S S\5r%g!\a SSKJr Npf=f)) annotationsN)Path) TYPE_CHECKINGAnyCallable)load_onnx_modelload_openvino_model)Self) AutoConfig AutoModel AutoTokenizer MT5ConfigPretrainedConfigT5Config)is_peft_available)find_adapter_config_file) InputModule PeftConfigc^^SUU4SjjnU$)Ncp>[R"[U5T- SS9 T"[U5T- 40UD6$)NT)exist_ok)osmakedirsr)save_directorykwargs_save_pretrained_fn subfolders b/home/james-whalen/.local/lib/python3.13/site-packages/sentence_transformers/models/Transformer.pywrapper)_save_pretrained_wrapper..wrappers5 D(94tD"4#7)#CNvNN)rz str | PathreturnNone)rrr s`` r_save_pretrained_wrapperr&sOO Nr"c0^\rSrSr%SrSrS\S'SS/rS\S 'S rS \S 'SSU4S jjjr SSjr SSjr SSjr SSjr S SjrS!SjrS"SjrS#S$SjjrS#S%Sjjr\S&S'Sjj5r\S&S(Sjj5r\S)S*U4Sjjj5rSrU=r$)+ Transformer%aHugging Face AutoModel to generate token embeddings. Loads the correct class, e.g. BERT / RoBERTa etc. Args: model_name_or_path: Hugging Face models name (https://huggingface.co/models) max_seq_length: Truncate any inputs longer than max_seq_length model_args: Keyword arguments passed to the Hugging Face Transformers model tokenizer_args: Keyword arguments passed to the Hugging Face Transformers tokenizer config_args: Keyword arguments passed to the Hugging Face Transformers config cache_dir: Cache dir for Hugging Face Transformers to store/load models do_lower_case: If true, lowercases the input (independent if the model is cased or not) tokenizer_name_or_path: Name or path of the tokenizer. When None, then model_name_or_path is used backend: Backend used for model inference. Can be `torch`, `onnx`, or `openvino`. Default is `torch`. sentence_bert_config.jsonstrconfig_file_namemax_seq_length do_lower_casez list[str] config_keysTbool save_in_rootc l>[T U]5 XplXlUc0nUc0nUc0nUR XX5upUR "XXiU 40UD6 [ [R"URR5R5n [U 51Sk-Ul Ub SU;aX$S'[R"UbUOU4SU0UD6UlUc[#URS5(a~[#URR$S5(aY[#UR S5(a>['URR$R(UR R*5nX lUb:UR R.R0URR$lgg)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 tokenizerhasattrr9minr:r7r- __class____name__tokenizer_class)selfmodel_name_or_pathr- model_argstokenizer_args config_argsr8r.tokenizer_name_or_pathr=r9 is_peft_modelrGrLs rr<Transformer.__init__As *  J  !N  K $ 1 12DQX f +YeZde $G$5$5doo6M6M$N$Y$YZ$'(<$=A % !  %*??!$T__%;%;%S%SUYUcUcUtUt!u, ! -59^^5M5M5V5VDOO " " 2 .r"c J[UUURS5URS5URSS5S9bJ[5(d [S5eUS:wa [ S5eS S KJn UR"U40UDS U0D6S 4$[R"U40UDS U0D6S4$) ahLoads 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)r8rXrYrZzgLoading 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.rrr8T) rgetr Exception ValueErrorpeftrrHr )rOrPr8r=rSrs rr>Transformer._load_configzs %"#!oog.$4!,1CU!K   %&&}'! w (--.@eKe[degkk k))*<a aW`achhhr"c US:XaU(aSHnURUS5 M [U[5(aUR"XU40UD6 g[U[5(aUR "XU40UD6 g[ R"U4X#S.UD6UlgUS:Xa[S UUSS.UD6UlgUS:Xa[S UUSS.UD6Ulg[S US 35e) 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. r[)rYNr9r8onnxzfeature-extraction)rPr9 task_nameopenvinozUnsupported backend 'z6'. `backend` should be `torch`, `onnx`, or `openvino`.r%) pop isinstancer_load_t5_modelr_load_mt5_modelr rHrCrr r^)rOrPr9r8r=rUrQadapter_only_kwargs rr?Transformer._load_models ( g *6&NN#5t<+7&(++##$6 XZXFI..$$%7YjY"+";";&#/5#NX# -#5. DO  "1#5. DO4WI=stu ur"c TSSKJn S/UlUR"U4X#S.UD6Ulg)Loads the encoder model from T5r)T5EncoderModel decoder.*rbN) transformersrn"_keys_to_ignore_on_load_unexpectedrHrC)rOrPr9r8rQrns rrhTransformer._load_t5_models4/=HM9(88  '- FP r"c TSSKJn S/UlUR"U4X#S.UD6Ulg)rmr)MT5EncoderModelrorbN)rprtrqrHrC)rOrPr9r8rQrts rriTransformer._load_mt5_models40>I]:)99  '- FP r"cvS[UR5URRRS9S3$)Nz Transformer() architecture))dictget_config_dictrCrLrMrOs r__repr__Transformer.__repr__s3d4#7#7#9HaHaHjHjkllmnnr"c UR5VVs0sHup4X0R;dMX4_M nnnUR"S 0UDUDSS0D6nUSnXqS'[5(aSSKJn [ URU5(aURRR(alURS5n USn [R"XRRRU RS9n [R"X4SS 9US'URRR (aS U;aUS US 'U$s snnf) 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)PeftModelForFeatureExtractionr5)device)dim hidden_statesall_layer_embeddingsr%)itemsrGrCrr_rrgactive_peft_configis_prompt_learningsizer[onesnum_virtual_tokensrcatr9output_hidden_states) rOfeaturesrkeyvaluetrans_featuresoutputsrr batch_sizer5prefix_attention_masks rrDTransformer.forwards.,8@~~7Gl7G3RkRkKk*#*7Gl//ONOfO$O"1:'7#$    :4??,IJJOO66II-2215 !)*:!;(-  B B U U^l^s^s)%.3YY8M7^de-f)* ?? ! ! 6 6?g;U/6/GH+ ,3ms E E cBURRR$)N)rCr9 hidden_sizer{s rget_word_embedding_dimension(Transformer.get_word_embedding_dimensions%%111r"c 0n[US[5(aU/nO[US[5(a]/n/US'UHLn[[ UR 555upgUR U5 USR U5 MN U/nO7//pUH+n UR U S5 U R U S5 M- X/nUV V s/sH,oV s/sHn [U 5R5PM sn PM. nn n UR(a1UV V s/sH"oV s/sHoR5PM sn PM$ nn n URUR"UUSSURS.65 U$s sn fs sn n fs sn fs sn n f)z-Tokenizes a text and maps tokens to token-idsr text_keysr longest_firstpt)padding truncationreturn_tensors max_length) rgr+rynextiterrappendstripr.lowerupdaterIr-) rOtextsroutput to_tokenizelookuptext_keytextbatch1batch2 text_tuplecolss rtokenizeTransformer.tokenize!sn eAh $ $ 'K a$ ' 'K"$F; !%d6<<>&:!;""4({#**84 '-KF#  jm, jm,$"*KALL 41A 4 L   ?JK{s3s!GGIs3{KK NN*#..    5L4Ks0 E=#E8 <E= F%F >F8E=Fc URRXS9 URRU5 URU5 g)N)safe_serialization)rCsave_pretrainedrI save_config)rO output_pathrrs rsaveTransformer.saveIs7 '' '[ &&{3 %r"c  HURUUUUUUUUU U U S9 n U"SSU0U D6$)N) rPrrX cache_folderrYrZtrust_remote_code model_kwargstokenizer_kwargs config_kwargsr=rPr%)_load_init_kwargs)clsrPrrXrrYrZrrrrr=r init_kwargss rloadTransformer.loadNsP$++1%-/%-',  H&8HKHHr"c URUUUUUUS9n UUUUUS.nSU ;a0U S'SU ;a0U S'SU ;a0U S'U SRU5 U SRU5 U SRU5 U(aU SRU5 U (aU SRU 5 U (aU SRU 5 0U EXKS.E$)N)rPrrXrrYrZ)rrXrYrZrrQrRrS)r8r=) load_configr)rrPrrXrrYrZrrrrr=rr9 hub_kwargss rrTransformer._load_init_kwargsos $1%- ! # 0!2   v %#%F< 6 )')F# $  &$&F= ! |##J/ '' 3}$$Z0  < ' ' 5  # $ + +,< =  = ! ( ( 7H&H|HHr"c @>U(aU/O/SQnUHn[T U]UUUUUUUS9n U (dM O SW ;aSU S;aU SRS5 SU ;aSU S;aU SRS5 SU ;aSU S;aU SRS5 U $)N)r*zsentence_roberta_config.jsonzsentence_distilbert_config.jsonzsentence_camembert_config.jsonzsentence_albert_config.jsonz sentence_xlm-roberta_config.jsonzsentence_xlnet_config.json)rPrconfig_filenamerXrrYrZrQrrRrS)r;rrf) rrPrrrXrrYrZconfig_filenamesr9rLs rrTransformer.load_configs   0OW(#5# /)!!1)Fv 0 6 !&9VL=Q&Q < $ $%8 9 v %*=HXAY*Y # $ ( ()< = F "':f]>S'S = ! % %&9 : r")rCr=r.r-rGrI)NNNNNFNr[)rPr+r-z int | NonerQdict[str, Any] | NonerRrrSrr8 str | Noner.r0rTrr=r+r#r$) rPr+r8rr=r+rSdict[str, Any]r#z*tuple[PeftConfig | PretrainedConfig, bool]) rPr+r9zPeftConfig | PretrainedConfigr8r+r=r+rUr0r#r$)rPr+r9rr8r+r#r$)r#r+)rdict[str, torch.Tensor]r#r)r#int)T)rz.list[str] | list[dict] | list[tuple[str, str]]rz str | boolr#r)rr+rr0r#r$) NNNFFNNNr[)rPr+rr+rXbool | str | NonerrrYrrZr0rr0rrrrrrr=r+r#r )rPr+rr+rXrrrrYrrZr0rr0rrrrrrr=r+r#r)rNNNNF)rPr+rr+rrrXrrrrYrrZr0r#r)rM __module__ __qualname____firstlineno____doc__r,__annotations__r/r1r<r>r?rhrir|rDrrr classmethodrrr__static_attributes__ __classcell__)rLs@rr(r(%si.8c7.@K@L$ &*,004-1 $#-17W7W#7W* 7W . 7W + 7W7W7W!+7W7W 7W7Wr(i"%(i2<(iGJ(iYg(i 3(iT2v2v.2v 2v  2v  2v 2vh  o/b2\`&C&NX& &P&  #'#'#!&"'.226/3II I ! I ! III I,I0I-II !II@ #'#'#!&"'.226/37I7I 7I ! 7I ! 7I7I7I 7I,7I07I-7I7I !7I7Ir&*#'#'#!&+++$ + ! + ! +++ ++r"r()rrrr+r#zCallable[..., None])& __future__rrAloggingrpathlibrtypingrrrsentence_transformers.backendrr r ImportErrortyping_extensionsr[rpr r r rrrtransformers.utils.import_utilsrtransformers.utils.peft_utilsr(sentence_transformers.models.InputModuler getLoggerrMloggerr_rr&r(r%r"rrs" //N' dd=B@   8 $&((p+p1'&'sB B#"B#