SSKJr SSKJr SSKJr SSKrSSKJr SSK J r SSK J r J r JrJr SSKJr SS KJr SS KJrJrJrJr \R2"\5r"S S \R85r\"S S\55r\"SS\55r\"SS\55r g))ABC)partial)OptionalN)Cache)BaseModelOutputWithPastQuestionAnsweringModelOutput SequenceClassifierOutputWithPastTokenClassifierOutput) AutoModel)Unpack)TransformersKwargsauto_docstringcan_return_tupleloggingc0^\rSrSrSrSrU4SjrSrU=r$)GradientCheckpointingLayer$aBase class for layers with gradient checkpointing. This class enables gradient checkpointing functionality for a layer. By default, gradient checkpointing is disabled (`gradient_checkpointing = False`). When `model.set_gradient_checkpointing()` is called, gradient checkpointing is enabled by setting `gradient_checkpointing = True` and assigning a checkpointing function to `_gradient_checkpointing_func`. Important: When using gradient checkpointing with `use_reentrant=True`, inputs that require gradients (e.g. hidden states) must be passed as positional arguments (`*args`) rather than keyword arguments to properly propagate gradients. Example: ```python >>> # Correct - hidden_states passed as positional arg >>> out = self.layer(hidden_states, attention_mask=attention_mask) >>> # Incorrect - hidden_states passed as keyword arg >>> out = self.layer(hidden_states=hidden_states, attention_mask=attention_mask) ``` Fc>UR(aUR(aSnURRnSUS3nSU;aUS(a SUS'US- nSnSU;aUSb SUS'US- nSnS U;aUS b SUS 'US - nSnS U;aUS b SUS 'US - nSnU(a)UR S 5S-n[ R U5 UR"[[TU](40UD6/UQ76$[TU]("U0UD6$)NFz7Caching is incompatible with gradient checkpointing in z . Setting use_cachez `use_cache=False`,Tpast_key_valuez `past_key_value=None`,past_key_valuesz `past_key_values=None`, layer_pastz `layer_past=None`,,.) gradient_checkpointingtraining __class____name__rstriploggerwarning_gradient_checkpointing_funcrsuper__call__)selfargskwargsdo_warn layer_namemessagers T/var/www/html/shao/venv/lib/python3.13/site-packages/transformers/modeling_layers.pyr%#GradientCheckpointingLayer.__call__=sI  & &4==G00JOPZ|[deGf$ )<&+{#00 6)f5E.F.R+/'(44 F*v6G/H/T,0()55v%&*>*J'+|$00!..-3w'44WUW=M5XQW5X`[_` `w000) r __module__ __qualname____firstlineno____doc__rr%__static_attributes__ __classcell__rs@r,rr$s,#!1!1r.rc^\rSrSrSrU4Sjr\\SS\\ RS\\ RS\\ RS\\ S\\ RS \\ RS \\S \\S \4S jj55rSrU=r$) GenericForSequenceClassificationamodelc>[TU]U5 URUl[XR[ R "U55 [R"URURSS9Ul UR5 g)NF)bias) r$__init__ num_labelssetattrbase_model_prefixr from_confignnLinear hidden_sizescore post_initr&configrs r,r=)GenericForSequenceClassification.__init__esb   ++,,i.C.CF.KLYYv114??O  r. input_idsattention_mask position_idsr inputs_embedslabelsrr(returnc [XR5"U4UUUUUS.UD6n U Rn URU 5n UbURSn OURSn UR R cU S:wa [S5eUR R cSn OUbXR R :gRU R[R5n[R"URSU R[RS9nX-RS5n O.Sn [RUR R"S35 U [R"XRS9U 4nSnUbUR%XUUR S 9n['UUU R(U R*U R,S 9$) NrKrLrrMrrrz=Cannot handle batch sizes > 1 if no padding token is defined.)devicedtypez will not detect padding tokens in `inputs_embeds`. Results may be unexpected if using padding tokens in conjunction with `inputs_embeds.`)rS)logitsrN pooled_logitsrH)lossrUr hidden_states attentions)getattrr@last_hidden_staterEshaperH pad_token_id ValueErrortorStorchint32arangeargmaxr! warning_oncerr loss_functionr rrXrY)r&rJrKrLrrMrNrr(transformer_outputsrXrU batch_sizelast_non_pad_token non_pad_mask token_indicesrVrWs r,forward(GenericForSequenceClassification.forwardos8?tE[E[7\ 8 )%+' 8 8 ,== M*  "+J&,,Q/J ;; # # + a\] ] ;; # # +!#   "%)A)AAEEfmmUZU`U`aL!LL)!F!Fr!J !#    >>**+,ZZ  u||J}}MOaab   %%VR_hlhshs%tD/ /??-;;*55   r.)r>rENNNNNNN)rr0r1r2r@r=rrrr` LongTensorTensorr FloatTensorboolr rr rkr4r5r6s@r,r8r8as151537+/59-1$(8 E,,-8 !.8 u//0 8 "% 8   1 12 8 ))*8 D>8 +,8  *8 8 r.r8c4^\rSrSrSrU4SjrSrSr\\ SS\ \ RS\ \ RS\ \ RS \ \S \ \ RS \ \ RS \ \ RS \\S\4Sjj55rSrU=r$)GenericForQuestionAnsweringr:c>[TU]U5 [XR[R "U55 [ R"URS5Ul UR5 g)N) r$r=r?r@r rArBrCrD qa_outputsrFrGs r,r=$GenericForQuestionAnswering.__init__sO  ,,i.C.CF.KL))F$6$6: r.c@[XR5R$NrZr@ embed_tokens)r&s r,get_input_embeddings0GenericForQuestionAnswering.get_input_embeddingsst334AAAr.c8U[XR5lgrzr{)r&values r,set_input_embeddings0GenericForQuestionAnswering.set_input_embeddingss=B,,-:r.rJrKrLrrMstart_positions end_positionsr(rOc [XR5"U4UUUUS.UD6n U Rn URU 5n U R SSS9upU R S5R 5n U R S5R 5n SnUbUbUR"XXg40UD6n[UU U U RU RS9$)N)rKrLrrMrrR)dim)rW start_logits end_logitsrXrY) rZr@r[rwsplitsqueeze contiguousrer rXrY)r&rJrKrLrrMrrr(outputssequence_outputrUrrrWs r,rk#GenericForQuestionAnswering.forwards,349O9O+P , )%+' ,  , "331#)<<r<#: #++B/::< ''+668   &=+D%%libhiD+%!!//))   r.)rwrm)rr0r1r2r@r=r}rrrrr`rnrorrpr rr rkr4r5r6s@r,rsrssBC151537+/596:48% E,,-% !.% u//0 % "% %   1 12 % "%"2"23%   0 01% +,%  &% % r.rsc ^\rSrSrSrU4Sjr\\SS\\ RS\\ RS\\ RS\\ S\\ RS \\ RS \\S \4S jj55rS rU=r$)GenericForTokenClassificationr:c>[TU]U5 URUl[XR[ R "U55 [USS5b URnO[USS5b URnOSn[R"U5Ul [R"URUR5UlUR!5 g)Nclassifier_dropouthidden_dropoutg?)r$r=r>r?r@r rArZrrrBDropoutdropoutrCrDrErF)r&rHrrs r,r=&GenericForTokenClassification.__init__s   ++,,i.C.CF.KL 6/ 6 B!'!:!:  V-t 4 @!'!6!6 !$ zz"45 YYv1163D3DE  r.rJrKrLrrMrNrrOc $[XR5"U4UUUUUS.UD6n U Rn URU 5n UR U 5n Sn UbUR XUR 5n [U U U RU RS9$)NrQ)rWrUrXrY) rZr@r[rrErerHr rXrY) r&rJrKrLrrMrNrr(rrrUrWs r,rk%GenericForTokenClassification.forwards,349O9O+P , )%+' , , "33,,7O,  %%fdkkBD$!//))   r.)rr>rErm)rr0r1r2r@r=rrrr`rnrorrprqr rkr4r5r6s@r,rrs"151537+/59-1$(! E,,-! !.! u//0 ! "% !   1 12 ! ))*! D>!  ! ! r.r)!abcr functoolsrtypingrr`torch.nnrB cache_utilsrmodeling_outputsrr r r models.autor processing_utilsr utilsrrrr get_loggerrr!Modulerr8rsrr/r.r,rs  #$PP   H %:1:1zG sG G T9 #9 9 x7 C7 7 r.