+ j`^RIt^RIt^RIHt^RIt^RIt^RIHt^RIH t ^RI H t ^RI Ht^RIHt^RIHt^R IHt^R IHt^R IHt^R IHt^R IHtHtHtHtH t H!t!^RI"H#t#H$t$^RI%H&t&^RI'H(t(H)t)H*t*^RI+H,t,H-t-^RI.H/t/H0t0^RI1H2t2!RR]4t3!RR]Ph4t5!RR]Ph4t6!RR]Ph4t7!RR]Ph4t8!RR]Ph4t9RDR R!llt:!R"R#]Ph4t;!R$R%]Ph4t<!R&R']4t=!R(R)]Ph4t>!R*R+]Ph4t?!R,R-]Ph4t@])!R.R/]$44tARER0R1lltB] tC])!R2R3]A44tD])!R4R57!R6R7]A44tE])!R8R57!R9R:]A44tF])!R;R<]A44tG!R=R>]Ph4tH!R?R@]Ph4tI])!RAR57!RBRC]A44tJ.RFOtKR#)GN)Callable)nn)CrossEntropyLoss)initialization)ACT2FN)filter_output_hidden_states)is_deepspeed_zero3_enabled)is_fsdp_managed_module)create_bidirectional_mask)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputCausalLMOutputSequenceClassifierOutputTokenClassifierOutputWav2Vec2BaseModelOutput XVectorOutput)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringis_peft_available)can_return_tuplemerge_with_config_defaults)OutputRecordercapture_outputsData2VecAudioConfigc<aa]tRt^6toRV3RlltRtRtVtV;t#)Data2VecAudioConvLayerc<\SV`4V^8dVPV^, ,M^VnVPV,Vn\ P !VPVPVPV,VPV,VPR7Vn \ P!VPRR7Vn \VP,VnR#)r) kernel_sizestridebiasTelementwise_affineN)super__init__conv_dim in_conv_dim out_conv_dimrConv1d conv_kernel conv_stride conv_biasconv LayerNorm layer_normrfeat_extract_activation activationselfconfiglayer_id __class__s&&&u/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/data2vec/modeling_data2vec_audio.pyr)Data2VecAudioConvLayer.__init__7s !#Car=cZVP^8dVRRRVP)13,pV#)rNNNNrUrCs&&r;rEData2VecAudioPadLayer.forwardVs6    ")!Q0F43F3F2F0F*FGMr=rZrHrPs@@r;rSrSQsKr=rSc8aa]tRt^\toV3RltRtRtVtV;t#) Data2VecAudioPositionalConvLayerc<\SV`4\P!VPVPVP VP ^,VP R7Vn\VP 4Vn \VP,Vn \P!VPRR7VnR#)r?)r#paddinggroupsFr&N)r(r)rr- hidden_sizeconv_pos_kernel_sizenum_conv_pos_embedding_groupsr1rSr_rr4r5r2r3r7r8r:s&&r;r))Data2VecAudioPositionalConvLayer.__init__]s II      33//1477   -V-H-HI  !?!?@,,v'9'9eTr=cVPV4pVPV4pVP^^4pVPV4pVP^^4pVP V4pV#)r1r_rBr3r5rCs&&r;rE(Data2VecAudioPositionalConvLayer.forwardlsd -0  ]3 %//15  6 %//15  6 r=)r5r1r3r_rHrPs@@r;r]r]\s Ur=r]c8aa]tRt^wtoV3RltRtRtVtV;t#)$Data2VecAudioPositionalConvEmbeddingc<\SV`4\P!\ VP 4Uu.uFp\ V4NK up4VnR#uupiN)r(r)r ModuleListrangerVr]layersr7r8_r:s&& r;r)-Data2VecAudioPositionalConvEmbedding.__init__xsF mm?DVEcEc?d e?d! -f 5?d e  esAcVP^^4pVPF pV!V4pK VP^^4pV#rg)rBrp)r7rDlayers&& r;rE,Data2VecAudioPositionalConvEmbedding.forward~sD%//15 [[E!-0M!%//15 r=)rprHrPs@@r;rkrkws r=rkcBaa]tRt^toRtV3RltRtRtRtVt V;t #)Data2VecAudioFeatureEncoderz.Construct the features from raw audio waveformc <\SV`4\P!\ VP 4Uu.uFp\ WR7NK up4VnRVnRVn R#uupi))r9FTN) r(r)rrnronum_feat_extract_layersr! conv_layersgradient_checkpointing_requires_grad)r7r8ir:s&& r;r)$Data2VecAudioFeatureEncoder.__init__s\ ==AFvGeGeAf gAfA #F 7Af g ',#" hsA(cPVP4F pRVnK RVnR#)FN) parameters requires_gradr}r7params& r;_freeze_parameters.Data2VecAudioFeatureEncoder._freeze_parameterss#__&E"'E '#r=cVR,pVP'dVP'dRVnVPF pV!V4pK V#)rYTrYN)r}trainingrr{)r7 input_valuesrD conv_layers&& r;rE#Data2VecAudioFeatureEncoder.forwardsL$W-     4===*.M '**J&}5M+r=)r}r{r|) rIrJrKrL__doc__r)rrErMrNrOrPs@@r;rxrxs8#$  r=rxc8aa]tRt^toV3RltRtRtVtV;t#)Data2VecAudioFeatureProjectioncL<\SV`4\P!VPR,VP R7Vn\P!VPR,VP4Vn \P!VP4Vn R#)rhepsNrA) r(r)rr2r*layer_norm_epsr3Linearra projectionDropoutfeat_proj_dropoutdropoutrds&&r;r)'Data2VecAudioFeatureProjection.__init__sf ,,vr':@U@UV))FOOB$79K9KLzz&":":; r=cnVPV4pVPV4pVPV4pW3#rm)r3rr)r7rDnorm_hidden_statess&& r;rE&Data2VecAudioFeatureProjection.forwards7!__];(:;  ]3 00r=)rr3rrHrPs@@r;rrs< 11r=rcV^8dQhR\PR\PR\PR\PR\PR,R\R,R\R \ \ ,/#) r?modulequerykeyvalueattention_maskNscalingrkwargs)rModuletorchTensorfloatrr)formats"r; __annotate__rs{%% II% <<% % << % LL4' % T\ %%'(%r=c VfVPR4R,p\P!WP^^44V,pVe W,p\P P VRR7p\P PWVPR7p\P!W4p V P^^4P4p W3#)Ndim)prrA) sizermatmulrBr functionalsoftmaxrr contiguous) rrrrrrrr attn_weights attn_outputs &&&&&&&, r;eager_attention_forwardrs**R.D(<<}}Q':;gEL!#4 ==((2(>L==((6??([L,,|3K''1-88:K  $$r=c\aa]tRt^toRtRV3RlV3RllltRV3RlRlltRtVtV;t #) Data2VecAudioAttentionz=Multi-headed attention from 'Attention Is All You Need' papercR<V^8dQhRS[RS[RS[RS[RS[RS[RS[R,/#) r? embed_dim num_headsr is_decoderr% is_causalr8N)intrboolr)rrQs"r;r#Data2VecAudioAttention.__annotate__s_CCCC C  C  CC$d*Cr=c<\SV`4WnW nW0nW,VnWpnVP V,VP8wd\RVP RV R24hVP R,VnW@n W`n \P!WVR7Vn \P!WVR7Vn\P!WVR7Vn\P!WVR7VnR#)z;embed_dim must be divisible by num_heads (got `embed_dim`: z and `num_heads`: z).)r%Nr)r(r)rrrhead_dimr8 ValueErrorrrrrrk_projv_projq_projout_proj) r7rrrrr%rr8r:s &&&&&&&&r;r)Data2VecAudioAttention.__init__s "" !.  MMI %$.. 8MdnnM]$YKr3 }}d* $"ii 4@ ii 4@ ii 4@  )TB r=c<V^8dQhRS[PRS[PR,RS[PR,RS[R,RS[S[,RS[S[PS[PR,3,/#)r?rDkey_value_statesNroutput_attentionsrreturn)rrrrr tuple)rrQs"r;rrs~0)0)||0) ,,-0) t+ 0)  $; 0)-.0) u||U\\D00 10)r=c VRJpVPRRp.VORNVPN5pVPV4PV4P ^^4p V'dTMTp .V PRRORNVPN5p VP V 4PV 4P ^^4p VP V 4PV 4P ^^4p \P!VPP\4pT!VV V V V3RVP'gRM VPRVPRV/VBwppVP!.VORN5!P!4pVP#V4pVV3#)z#Input shape: Batch x Time x ChannelNrrrrA)shaperrviewrBrrr get_interfacer8_attn_implementationrrrrreshaperr)r7rDrrrris_cross_attention input_shape hidden_shape query_statescurrent_stateskv_shape key_states value_statesattention_interfacerrs&&&&&, r;rEData2VecAudioAttention.forwards.T9$))#2. 88b8$--8 {{=166|DNNqRST -?)]B^))#2.BBDMMB[[055h?II!QO {{>277AKKAqQ (?(M(M KK , ,.E) %8      % $}}}C$,, % LL % 0 %  % ! \"));;;;FFH mmK0 L((r=) r8rrrrrrrrrrr)rFTFN)NNF) rIrJrKrLrr)rErMrNrOrPs@@r;rrs#GCC>0)0)0)r=rc8aa]tRtRtoV3RltRtRtVtV;t#)Data2VecAudioFeedForwardi#c<\SV`4\P!VP4Vn\P !VPVP4Vn \VP\4'd\VP,VnMVPVn\P !VPVP4Vn\P!VP 4VnR#rm)r(r)rractivation_dropoutintermediate_dropoutrraintermediate_sizeintermediate_dense isinstance hidden_actstrrintermediate_act_fn output_densehidden_dropoutoutput_dropoutrds&&r;r)!Data2VecAudioFeedForward.__init__$s $&JJv/H/H$I!"$))F,>,>@X@X"Y f'' - -'-f.?.?'@D $'-'8'8D $IIf&>&>@R@RS jj)>)>?r=cVPV4pVPV4pVPV4pVPV4pVP V4pV#rm)rrrrrrCs&&r;rE Data2VecAudioFeedForward.forward1sX// > 00? 11-@ ))-8 ++M: r=)rrrrrrHrPs@@r;rr#s @r=rcHaa]tRtRtoV3RltRV3RlRlltRtVtV;t#)Data2VecAudioEncoderLayeri;c<\SV`4\VPVPVP RVR7Vn\P!VP4Vn \P!VPVPR7Vn \V4Vn\P!VPVPR7VnR#)F)rrrrr8rN)r(r)rranum_attention_headsattention_dropout attentionrrrrr2rr3r feed_forwardfinal_layer_normrds&&r;r)"Data2VecAudioEncoderLayer.__init__<s /((00,,  zz&"7"78 ,,v'9'9v?T?TU4V< " V-?-?VEZEZ [r=cr<V^8dQhRS[PRS[PR,RS[S[,/#r?rDrNr)rrrr)rrQs"r;r&Data2VecAudioEncoderLayer.__annotate__Ks<|| t++, r=c TpVP!V3RV/VBwrVPV4pWA,pVPV4pWPV4,pVP V4pV#)r)rrr3rr)r7rDrr attn_residualrrs&&&, r;rE!Data2VecAudioEncoderLayer.forwardKsr & >>-aaZ`a  ]3 %5  6 %(9(9-(HH --m< r=)rrrrr3rmrHrPs@@r;rr;s \r=rcHaa]tRtRtoV3RltRV3RlRlltRtVtV;t#)Data2VecAudioEncoderi]c<\SV`4Wn\V4Vn\ P !VPVPR7Vn \ P!VP4Vn \ P!\VP4Uu.uFp\!V4NK up4VnRVnR#uupi)rFN)r(r)r8rkpos_conv_embedrr2rarr3rrrrnronum_hidden_layersrrpr|rqs&& r;r)Data2VecAudioEncoder.__init__^s  B6J,,v'9'9v?T?TUzz&"7"78 mmPUV\VnVnPo$pPo1%>v%FPo$pq &+#%qs#C cr<V^8dQhRS[PRS[PR,RS[S[,/#r)rtensorrrr)rrQs"r;r!Data2VecAudioEncoder.__annotate__gsA!@!@||!@ t+!@+, !@r=c Ve9VPR4P^^VP^,4p^W(&\VPVVR7pVP V4pWP VP4,pVPV4pVPV4p\4;'g \V4pVPFbp\P!.4pVP;'dWPP 8p V 'd V'gKWV!V3RV/VBpKd \#VR7#)N)r8 inputs_embedsrr)last_hidden_staterA) unsqueezerepeatrr r8rtodevicer3rr r rprrandr layerdropr) r7rDrrexpand_attention_maskposition_embeddings synced_gpusrudropout_probabilityskip_the_layers &&&, r;rEData2VecAudioEncoder.forwardgs  %$2$<$(>}?S?S(TT  6  ]3 02RR6LT6R [[E"'**R. !]]ZZ/B[[EZEZ/ZN![[ %m ]N ]V\ ] !??r=)r8rr|r3rprrmrHrPs@@r;rr]s,!@!@!@r=rc8aa]tRtRtoV3RltRtRtVtV;t#)Data2VecAudioAdapterLayeric<\SV`4\P!VP^VP,VP VP ^R7VnR#)r?)r$r_N)r(r)rr-output_hidden_sizeadapter_kernel_sizeadapter_strider1rds&&r;r)"Data2VecAudioAdapterLayer.__init__sJ II  % % )) )  & &((   r=cjVPV4p\PPV^R7pV#)rhr)r1rrglurCs&&r;rE!Data2VecAudioAdapterLayer.forwards/ -0  ))-Q)? r=)r1rHrPs@@r;rrs r=rc8aa]tRtRtoV3RltRtRtVtV;t#)Data2VecAudioAdapteric<a\SV`4SPSP8wdW\P !SPSP4Vn\P!SP4VnMR;VnVn\P!V3Rl\SP444Vn SPVn R#)Nc3:<"TFp\S4xK R#5irm)r).0rrr8s& r; 0Data2VecAudioAdapter.__init__..s#pOo!$=f$E$EOos)r(r)rrarrprojr2proj_layer_normrnronum_adapter_layersrprrds&fr;r)Data2VecAudioAdapter.__init__s   $ $(:(: : &"4"4f6O6OPDI#%<<0I0I#JD /3 3DI,mm#puU[UnUnOo#pp ))r=cVPe1VPe#VPV4pVPV4pVP^^4pVPFMp\P P 4pVP 'dW0P8gKEV!V4pKO VP^^4pV#rm)r)r*rBrpnprandomrr)r7rDrulayerdrop_probs&& r;rEData2VecAudioAdapter.forwards 99 T%9%9%E IIm4M 00?M%//15 [[EYY--/N===^nn%D %m 4 ! &//15 r=)rrpr)r*rHrPs@@r;r#r#s *r=r#c aa]tRtRtoRtRtRtRtRtRt Rt R] R] !] ^RR 7/t]P !4V3R l4tRV3R lR lltRV3R lRlltV3RltRtVtV;t#)Data2VecAudioPreTrainedModelidata2vec_audioraudioTrD attentionsencoder)index layer_namec t<\SV`V4\V\4'd\P !^VP P, 4p\P!VP PV)VR7\P!VP PV)VR7R#\V\4'd.\P!VPP^4R#\V\P 4'd\P"!VP4VPep\P !VP$VP&VP(^,,, 4p\P!VPV)VR7R#R#R#)zInitialize the weights)abN)r( _init_weightsrrmathsqrtr in_featuresinituniform_weightr%r] constant_r1rr-kaiming_normal_r` in_channelsr#)r7rkr:s&& r;r=*Data2VecAudioPreTrainedModel._init_weightss f% f< = = !f//;;; MM&++00QB! <  @ A A NN6;;++Q /  * *   /{{&IIfmmv/A/AFDVDVWXDY/YZ[ fkkaR15'+r=cX<V^8dQhRS[PS[,RS[R,/#)r? input_lengths add_adapterN)r LongTensorrr)rrQs"r;r)Data2VecAudioPreTrainedModel.__annotate__s.e>N>NQT>Tcgjncnr=c hVfVPPMTpRp\VPPVPP4FwrEV!WV4pK V'dE\ VPP 4F!pV!V^VPP4pK# V#)z8 Computes the output length of the convolutional layers cN\P!W, VRR7^,#)floor) rounding_mode)rdiv input_lengthr#r$s&&&r;_conv_out_lengthWData2VecAudioPreTrainedModel._get_feat_extract_output_lengths.._conv_out_lengths 99\7wWZ[[ [r=)r8rKzipr.r/ror+r)r7rJrKrUr#r$rrs&&& r; _get_feat_extract_output_lengths=Data2VecAudioPreTrainedModel._get_feat_extract_output_lengthss 2=1Ddkk--+  \ $'t{{'>'> @W@W#X K,]PM$Y 4;;99: 04;;C]C] ^ ;r=c:<V^8dQhRS[RS[P/#)r?feature_vector_lengthr)rrrL)rrQs"r;rrMs#%(:?:J:Jr=cVPRR7R,pVPWCR7pVP\P4pVP ^,p\P !Wa3VPVPR7p^V\P!VP ^,VPR7V^, 3&VPR.4PR4PR.4P4pV#)rhrrK)dtyper)rrA)rYrA) cumsumrXrrlongrzerosr^rarangeflipr)r7r[rrKnon_padded_lengthsoutput_lengths batch_sizes&&&& r;"_get_feature_vector_attention_mask?Data2VecAudioPreTrainedModel._get_feature_vector_attention_masks ,22r2:5A>>?Q>k'**5::6#))!,   /~7K7KTbTiTi uv ^%9%9!%<^EZEZ[]kno]opq',,bT299"=BBB4HMMOr=c&<V^8dQh/S[;R&#r?r8r)rrQs"r;rrMs r=rm)rIrJrKrLbase_model_prefixmain_input_nameinput_modalitiessupports_gradient_checkpointing_supports_flash_attn_supports_sdpa_supports_flex_attnrrr_can_record_outputsrno_gradr=rXrg__annotate_func__rMrNrOrPs@@r;r3r3s)$O&*#N2n%;1QZ[  ]]_66 *kr=r3c V^8dQhR\\\3,R\R\R\PR,R\R\ P /#)r?r mask_prob mask_lengthrN min_masksr)rrrrrLr.ndarray)rs"r;rrsctt c?ttt$$t+ t  t ZZ tr=c ^aaaaaVwpoS^8d \R4hSS8d\RS RS R24h\PP^4P 4oVVVVV3RlpVe.VP 4P R 4P4M\V4Uu.uFpSNK upp\P!VS3\R7p .p V!S4p V ^8XdV #VFp V!V 4p \PP\P!V S^, , 4V RR7p\V4^8Xd S^, pM V^,p\P!V\P!W, \P R7V,.4pV P#V4K \P$!V 4p \P&!V R ,W[S34p V P)W[S,4p \P!S4R ,p\P&!VW[S34P)W[S,4pV V,p V P+4S^, 8dS^, WS^, 8&\P,!W^R 4V #uupi) a2 Computes random mask spans for a given shape. Used to implement [SpecAugment: A Simple Data Augmentation Method for ASR](https://huggingface.co/papers/1904.08779). Note that this method is not optimized to run on TPU and should be run on CPU as part of the preprocessing during training. Args: shape: The shape for which to compute masks. This should be of a tuple of size 2 where the first element is the batch size and the second element is the length of the axis to span. mask_prob: The percentage of the whole axis (between 0 and 1) which will be masked. The number of independently generated mask spans of length `mask_length` is computed by `mask_prob*shape[1]/mask_length`. Note that due to overlaps, `mask_prob` is an upper bound and the actual percentage will be smaller. mask_length: size of the mask min_masks: minimum number of masked spans attention_mask: A (right-padded) attention mask which independently shortens the feature axis of each batch dimension. z&`mask_length` has to be bigger than 0.zO`mask_length` has to be smaller than `sequence_length`, but got `mask_length`: z and `sequence_length`: `c<\SV,S, S,4p\VS4pVS,S8d SS,pVS^, , V8d\VS^, , ^4pV#)z;Given input length, compute how many spans should be masked)rmax)rTnum_masked_spanepsilonrxrwrysequence_lengths& r;compute_num_masked_span6_compute_mask_indices..compute_num_masked_span,soi,6DwNOoy9 [ (? :-G[G[chcmcm n /3/E/E/H/HI\I\/]M , ;; ( (1 ,#8)++77 KK;;++<< $ $)<<0DMaMainisis#t #7#@#G#GO]_#` 23M /r=F)tie_last_hidden_statesc <V^8dQhRS[PR,RS[PR,RS[PR,RS[S[,RS[S[,/#)r?rNrrrr)rrrrrrData2VecAudioBaseModelOutput)rrQs"r;rrsf( ( llT)(  t+( !,,t3 ( +, ( - - ( r=c VPV4pVP^^4pVe&VPVP^,VRR7pVP V4wreVP WcVR7pVP !V3RV/VBpVPpVPeVPV4p\VVR7#)a mask_time_indices (`torch.BoolTensor` of shape `(batch_size, sequence_length)`, *optional*): Indices to mask extracted features for contrastive loss. When in training mode, model learns to predict masked extracted features in *config.proj_codevector_dim* space. Fr])rrr)r extract_features) rrBrgrrrr7r rr)r7rrrrrrDencoder_outputss&&&&, r;rEData2VecAudioModel.forwards 11,?+55a;  %!DD &&q)>uEN+/*A*ABR*S' 00 ~1 ,,  )   (99 << # LL7M++-  r=)rr8r7rrrNN)rIrJrKrLr)rrrrrrErMrNrOrPs@@r;rrsR"4,,\ E2( ( 3 ( ( r=rzu Data2VecAudio Model with a `language modeling` head on top for Connectionist Temporal Classification (CTC). ) custom_introcbaa]tRtRtoV3RltRt]]RV3RlRll44tRt Vt V;t #)Data2VecAudioForCTCic <\SV`V4\V4Vn\P !VP 4VnVPf\RVP R24h\VR4'dVP'd VPM VPp\P!W!P4VnVP#4R#)a2 config ([`Data2VecAudioForCTC`]): Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights. NzYou are trying to instantiate z with a configuration that does not define the vocabulary size of the language model head. Please instantiate the model as follows: `Data2VecAudioForCTC.from_pretrained(..., vocab_size=vocab_size)`. or define `vocab_size` of your model's configuration.rK)r(r)rr4rr final_dropoutr vocab_sizerr:hasattrrKrrarlm_headr)r7r8rr:s&& r;r)Data2VecAudioForCTC.__init__s  08zz&"6"67    $00@AHH *1)G)GFL^L^L^F % %djdvdv yy!35F5FG  r=c NVPPP4R#rr4rrrs&r;r*Data2VecAudioForCTC.freeze_feature_encoder --@@Br=c <V^8dQhRS[PR,RS[PR,RS[PR,RS[S[,RS[S[,/#r?rNrlabelsrr)rrrrrr)rrQs"r;r Data2VecAudioForCTC.__annotate__ sc: : llT):  t+:  t# : +, :   : r=c xVeLVP4VPP8d#\RVPP 24hVP!V3RV/VBpVP pVP V4pVPV4pRpVEenVeTM%\P!V\PR7pVPVPR 44P\P4p V^8p V PR 4p VPV 4p \P P#VR \P$R7P'^^4p \P(P*P-RR7;_uu_4\P P/V V V V VPP0VPP2VPP4R7pRRR4\7WVP8VP:R 7# +'giL2;i) a labels (`torch.LongTensor` of shape `(batch_size, target_length)`, *optional*): Labels for connectionist temporal classification. Note that `target_length` has to be smaller or equal to the sequence length of the output logits. Indices are selected in `[-100, 0, ..., config.vocab_size - 1]`. All labels set to `-100` are ignored (masked), the loss is only computed for labels in `[0, ..., config.vocab_size - 1]`. Nz$Label values must be <= vocab_size: rr)rr^F)enabled)blank reduction zero_infinitylosslogitsrDr6rA)r~r8rrr4r rrr ones_liker`rXrr masked_selectrr log_softmaxfloat32rBbackendscudnnflagsctc_loss pad_token_idctc_loss_reductionctc_zero_infinityrrDr6)r7rrrroutputsrDrrrJ labels_masktarget_lengthsflattened_targets log_probss&&&&, r;rEData2VecAudioForCTC.forwards  &**,$++2H2H"HCDKKDZDZC[\] ]%%  )    11  ]3 m,  #1"<%//R^fkfpfpBq !AA.BTBTUWBXY\\]b]g]ghM!A+K(__R0N & 4 4[ A  11&b 1V``abdefI%%++E+::}}--%!"++22"kk<<"&++"?"?.;G4I4IV]VhVh  ;:s A#H)) H9 )r4rrr) rIrJrKrLr)rrrrErMrNrOrPs@@r;rrs6 6C: : : : r=rz Data2VecAudio Model with a sequence classification head on top (a linear layer over the pooled output) for tasks like SUPERB Keyword Spotting. craa]tRtRtoV3RltRtRt]]] RV3RlRll444t Rt Vt V;t #) &Data2VecAudioForSequenceClassificationi]c:<\SV`V4\VR4'dVP'd \ R4h\ V4VnVP^,pVP'd7\P!\P!V4V, 4Vn \P!VPVP 4Vn\P!VP VP$4VnVP)4R#)rKzdSequence classification does not support the use of Data2VecAudio adapters (config.add_adapter=True)N)r(r)rrKrrr4ruse_weighted_layer_sumrrrr layer_weightsrraclassifier_proj_size projector num_labels classifierrr7r8 num_layersr:s&& r;r)/Data2VecAudioForSequenceClassification.__init__ds  6= ) )f.@.@.@v 18--1  ( ( (!#ejj.Dz.Q!RD 6#5#5v7R7RS))F$?$?ARARS r=c NVPPP4R#rrrs&r;r=Data2VecAudioForSequenceClassification.freeze_feature_encoderurr=c VVPP4F pRVnK R#z Calling this function will disable the gradient computation for the base model so that its parameters will not be updated during training. Only the classification head will be updated. FNr4rrrs& r;freeze_base_model8Data2VecAudioForSequenceClassification.freeze_base_model|% ((335E"'E 6r=c <V^8dQhRS[PR,RS[PR,RS[PR,RS[S[,RS[S[,/#r)rrrrrr)rrQs"r;r3Data2VecAudioForSequenceClassification.__annotate__sd9 9 llT)9  t+9  t# 9 +, 9 ) ) 9 r=c (VPP'dRVR&VP!V3RV/VBpVPP'dzVPp\P !V^R7p\ PPVPRR7pWgPR^^4,P^R7pM VPpVPV4pVfVP^R7pMVPVP ^,V4p V P#R4P%^^VP ^,4p RWj(&VP^R7V P^R7PR^4, pVP'V4p Rp VeG\)4p V !V PRVPP*4VPR44p \-V V VPVP.R7#)  input_values (`torch.FloatTensor` of shape `(batch_size, sequence_length)`): Float values of input raw speech waveform. Values can be obtained by loading a `.flac` or `.wav` audio file into an array of type `list[float]`, a `numpy.ndarray` or a `torch.Tensor`, *e.g.* via the torchcodec library (`pip install torchcodec`) or the soundfile library (`pip install soundfile`). To prepare the array into `input_values`, the [`AutoProcessor`] should be used for padding and conversion into a tensor of type `torch.FloatTensor`. See [`Data2VecAudioProcessor.__call__`] for details. labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If `config.num_labels > 1` a classification loss is computed (Cross-Entropy). Toutput_hidden_statesrrNrrrA)r8rr4rDrstackrrrrrrr rmeanrgrr r rrrrr6)r7rrrrrrD norm_weights pooled_output padding_maskexpand_padding_maskrrloss_fcts&&&&, r;rE.Data2VecAudioForSequenceClassification.forwards, ;; - - --1F) *%%  )   ;; - - -#11M!KK 1=M==001C1C0LL*->->r1a-HHMMRSMTM#55M}5  !)..1.5MBB=CVCVWXCY[ijL"."8"8"<"C"CAq-J]J]^_J`"a 25M. /)--!-4|7G7GA7G7N7S7STVXY7ZZM/  ')HFKKDKK,B,BCV[[QS_UD'!//))   r=)rr4rrrrIrJrKrLr)rrrrrrErMrNrOrPs@@r;rr]sC"C( 9 9 !9 9 r=rcraa]tRtRtoV3RltRtRt]]] RV3RlRll444t Rt Vt V;t #) (Data2VecAudioForAudioFrameClassificationic<\SV`V4\VR4'dVP'd \ R4h\ V4VnVP^,pVP'd7\P!\P!V4V, 4Vn \P!VPVP 4VnVP VnVP%4R#)rKzgAudio frame classification does not support the use of Data2VecAudio adapters (config.add_adapter=True)N)r(r)rrKrrr4rrrrrrrrrarrrrs&& r;r)1Data2VecAudioForAudioFrameClassification.__init__s  6= ) )f.@.@.@y 18--1  ( ( (!#ejj.Dz.Q!RD ))F$6$68I8IJ ++ r=c NVPPP4R#rrrs&r;r?Data2VecAudioForAudioFrameClassification.freeze_feature_encoderrr=c VVPP4F pRVnK R#rrrs& r;r:Data2VecAudioForAudioFrameClassification.freeze_base_modelrr=c <V^8dQhRS[PR,RS[PR,RS[PR,RS[S[,RS[S[,/#r)rrrrrr)rrQs"r;r5Data2VecAudioForAudioFrameClassification.__annotate__sd0 0 llT)0  t+0  t# 0 +, 0 & & 0 r=c VPP'dRVR&VP!V3RV/VBpVPP'dzVPp\P !V^R7p\ PPVPRR7pWgPR^^4,P^R7pM VPpVPV4pRp Ve^\4p V !VPRVP4\P !VPRVP4^R74p \#V VVPVP$R7#) rTrrrN)axisrrA)r8rr4rDrrrrrrrrr rrrargmaxrr6) r7rrrrrrDrrrr s &&&&, r;rE0Data2VecAudioForAudioFrameClassification.forwardsA, ;; - - --1F) *%%  )   ;; - - -#11M!KK 1=M==001C1C0LL*->->r1a-HHMMRSMTM#55M/  ')HFKKDOOtuD$!//))   r=)rr4rrrrrPs@@r;rrsC C( 0 0 !0 0 r=rc<aa]tRtRtoRV3RlltRtRtVtV;t#) AMSoftmaxLossic<\SV`4W0nW@nW n\ P !\P!W4RR7Vn \ P!4Vn R#)T)rN) r(r)scalemarginrrrrrandnrCrr)r7 input_dimrr r!r:s&&&&&r;r)AMSoftmaxLoss.__init__sI   $ll5;;y#EUYZ '') r=cVP4p\PPVP^R7p\PPV^R7p\ P !W4pW@P, p\PPW P4pVP\ P!VP4WT4,pVPWr4pV#)rr)flattenrr normalizerCrmmr!one_hotrr whererr) r7rDrrC cos_thetapsionehotrrs &&& r;rEAMSoftmaxLoss.forward#s!((!(< // 1/E HH]3 ++%&&v?ekk&++-HHyy( r=)rr!rr rC)g>@g?rHrPs@@r;rrs*  r=rcHaa]tRtRtoRV3RlltV3RlRltRtVtV;t#) TDNNLayeri1c<\SV`4V^8dVPV^, ,MVPV,VnVPV,VnVP V,VnVPV,Vn\P!VPVP ,VP4Vn \P!4Vn R#)rN)r(r)tdnn_dimr+r, tdnn_kernelr# tdnn_dilationdilationrrkernelReLUr5r6s&&&r;r)TDNNLayer.__init__2s +BFD]D]!^))F$=$=v?X?XY&v'@'@&BSBST RsF)c NVPPP4R#rrrs&r;r.Data2VecAudioForXVector.freeze_feature_encoderjrr=c VVPP4F pRVnK R#rrrs& r;r)Data2VecAudioForXVector.freeze_base_modelqrr=cD<V^8dQhRS[PS[,/#)r?rJ)rrLr)rrQs"r;r$Data2VecAudioForXVector.__annotate__ys  e6F6F6L r=c XRpVPPF pV!W^4pK V#)z/ Computes the output length of the TDNN layers c.W, V,^,#rgrkrSs&&&r;rUJData2VecAudioForXVector._get_tdnn_output_lengths.._conv_out_length~s!.69A= =r=)r8r3)r7rJrUr#s&& r;_get_tdnn_output_lengths0Data2VecAudioForXVector._get_tdnn_output_lengthsys1  >  ;;22K,]KM3r=c <V^8dQhRS[PR,RS[PR,RS[PR,RS[S[,RS[S[,/#r)rrrrrr)rrQs"r;rrOsjF F llT)F  t+F  t# F +, F   F r=c VPP'dRVR&VP!V3RV/VBpVPP'dzVPp\P !V^R7p\ PPVPRR7pWgPR^^4,P^R7pM VPpVPV4pVPF pV!V4pK Vf&VP^R7p VP!^R7p MVP#VP^R74p VP%V 4p .p .p \'V 4F[wrV P)WmRV13,P^R74V P)WmRV13,P!^R74K] \P !V 4p \P !V 4p \P*!W.RR7pVP-V4pVP/V4pRpVeVP1VV4p\3VVVVPVP4R7#)rTrrrN)rr embeddingsrDr6rA)r8rr4rDrrrrrrrrr rrErstdrXrS enumeratercatrrrGrr6)r7rrrrrrDr tdnn_layer mean_features std_featuresfeat_extract_output_lengthstdnn_output_lengthsr~lengthstatistic_poolingoutput_embeddingsrrs&&&&, r;rEData2VecAudioForXVector.forwardsH, ;; - - --1F) *%%  )   ;; - - -#11M!KK 1=M==001C1C0LL*->->r1a-HHMMRSMTM#55M}5 ))J&}5M$  !)..1.5M(,,,3L*.*O*OP^PbPbghPbPi*j '"&"?"?@["\ ML&':; $$]gvg:%>%C%C%C%JK##MWfW*$=$A$Aa$A$HI<"KK 6M ;;|4L!II}&CL 223DE!23  >>&&1D(!//))   r=)rr4rrrGrrEr)rIrJrKrLr)rrrSrrrrErMrNrOrPs@@r;rCrCQsQ &C(   F F !F F r=rC)Nr)Nr)rrrrCrr3)Lr>r>collections.abcrnumpyr.rrtorch.nnrrrA activationsrbackbone_utilsrintegrations.deepspeedr integrations.fsdpr masking_utilsr modeling_flash_attention_utilsr modeling_layersr modeling_outputsrrrrrrmodeling_utilsrrprocessing_utilsrutilsrrr utils.genericrrutils.output_capturingrrconfiguration_data2vec_audiorr!rrSr]rkrxrrrrrrrr#r3rrrrrrrr0rC__all__rkr=r;rws0* $ %&!9@76B9G&JJIE=76BIIryy6 299 ")): 1RYY 1%8R)RYYR)jryy0 :D+@299+@\ $299>F?FFRtn 7r 5r r j _ 6_  _ D ] -I]  ] @S /KS S lBII. @ { :{  { | r=