U ,-e@sdZddlZddlmZmZmZddlZddlZddl Zddlm Z ddl m Z ddl mZddlmZdd lmZmZmZdd lmZdd lmZmZmZmZmZd d lmZee Z!d Z"dZ#dZ$d ddgZ%dZ&dZ'dZ(dZ)dZ*dgZ+dFee,e,fe-e,eej.e,ej/dddZ0Gddde j1Z2Gddde j1Z3Gdd d e j1Z4Gd!d"d"e j1Z5Gd#d$d$e j1Z6Gd%d&d&e j1Z7Gd'd(d(e7Z8Gd)d*d*e j1Z9Gd+d,d,e j1Z:Gd-d.d.e j1Z;Gd/d0d0e j1ZGd5d6d6e j1Z?Gd7d8d8e j1Z@Gd9d:d:eZAd;ZBdd?d?eAZDed@eBGdAdBdBeAZEedCeBGdDdEdEeAZFdS)Gz PyTorch Hubert model.N)OptionalTupleUnion)nn)CrossEntropyLoss)ACT2FN)is_deepspeed_zero3_enabled)BaseModelOutputCausalLMOutputSequenceClassifierOutput)PreTrainedModel)add_code_sample_docstringsadd_start_docstrings%add_start_docstrings_to_model_forwardloggingreplace_return_docstrings) HubertConfigrzfacebook/hubert-large-ls960-fti$iz['MISTER QUILTER IS THE APOSTLE OF THE MIDDLE CLASSES AND WE ARE GLAD TO WELCOME HIS GOSPEL'gGz6@zsuperb/hubert-base-superb-ksz '_unknown_'g(\!@zfacebook/hubert-base-ls960)shape mask_prob mask_lengthattention_mask min_masksreturncs|\}dkrtdkr6tdddtjdfdd}|dk rt|d nfd d t|D}tj |ft d }g} |} | d kr|S|D]v} || } tjj t | d| dd} t | d krd}n| d }t| tj| | tjd |g} | | qt| } t| dddddf|| f} | || } t ddddf}t||| f|| }| |} | dkr҈d| | dk<t|| dd |S)af Computes random mask spans for a given shape. Used to implement [SpecAugment: A Simple Data Augmentation Method for ASR](https://arxiv.org/abs/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. rz&`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`: `csXt|}t|}|kr2}|d|krTt|dd}|S)z;Given input length, compute how many spans should be maskedrr)intmax) input_lengthnum_masked_spanepsilonrrrsequence_lengthk/var/www/html/Darija-Ai-Train/env/lib/python3.8/site-packages/transformers/models/hubert/modeling_hubert.pycompute_num_masked_spanjs  z6_compute_mask_indices..compute_num_masked_spanNcsg|]}qSr#r#.0_)r"r#r$ }sz)_compute_mask_indices..dtyperF)replace) ValueErrornprandomranditemsumdetachtolistrangezerosboolchoicearangelenZ concatenateonesZint32appendarrayZ broadcast_toreshaperZput_along_axis)rrrrr batch_sizer% input_lengthsZ spec_aug_maskZspec_aug_mask_idxsZmax_num_masked_spanrrZspec_aug_mask_idxZdummy_mask_idxoffsetsr#r r$_compute_mask_indicesDs`    rCcs&eZdZdfdd ZddZZS)HubertNoLayerNormConvLayerrcsjt|dkr |j|dnd|_|j||_tj|j|j|j||j||j d|_ t |j |_ dS)Nrr kernel_sizestridebias)super__init__conv_dim in_conv_dim out_conv_dimrConv1d conv_kernel conv_stride conv_biasconvrfeat_extract_activation activationselfconfiglayer_id __class__r#r$rJs  z#HubertNoLayerNormConvLayer.__init__cCs||}||}|SN)rRrTrV hidden_statesr#r#r$forwards  z"HubertNoLayerNormConvLayer.forward)r__name__ __module__ __qualname__rJr^ __classcell__r#r#rYr$rDsrDcs&eZdZdfdd ZddZZS)HubertLayerNormConvLayerrcs|t|dkr |j|dnd|_|j||_tj|j|j|j||j||j d|_ tj |jdd|_ t |j|_dS)NrrrET)Zelementwise_affine)rIrJrKrLrMrrNrOrPrQrR LayerNorm layer_normrrSrTrUrYr#r$rJs  z!HubertLayerNormConvLayer.__init__cCs:||}|dd}||}|dd}||}|S)Nr&)rR transposerfrTr\r#r#r$r^s      z HubertLayerNormConvLayer.forward)rr_r#r#rYr$rdsrdcs&eZdZdfdd ZddZZS)HubertGroupNormConvLayerrcst|dkr |j|dnd|_|j||_tj|j|j|j||j||j d|_ t |j |_ tj|j|jdd|_dS)NrrrET)Z num_groupsZ num_channelsZaffine)rIrJrKrLrMrrNrOrPrQrRrrSrT GroupNormrfrUrYr#r$rJs   z!HubertGroupNormConvLayer.__init__cCs"||}||}||}|Sr[)rRrfrTr\r#r#r$r^s   z HubertGroupNormConvLayer.forward)rr_r#r#rYr$risrics$eZdZfddZddZZS)HubertPositionalConvEmbeddingc sttj|j|j|j|jd|jd|_tjj }t tjj drNtjj j }t rddl }|jj|jjdd||jddd|_W5QRX|j||jj|j||jjn||jddd|_t|j|_t|j|_dS)N)rFpaddinggroups weight_normrZ modifier_rankweight)namedim)rIrJrrN hidden_sizenum_conv_pos_embeddingsZnum_conv_pos_embedding_groupsrRutilsrohasattrZparametrizationsr deepspeedzeroGatheredParametersrqZregister_external_parameterweight_vweight_gHubertSamePadLayerrmrrSrT)rVrWrorxrYr#r$rJs(   z&HubertPositionalConvEmbedding.__init__cCs:|dd}||}||}||}|dd}|SNrrl)rhrRrmrTr\r#r#r$r^#s      z%HubertPositionalConvEmbedding.forwardr_r#r#rYr$rks rkcs$eZdZfddZddZZS)r}cs$t|ddkrdnd|_dS)Nrlrr)rIrJnum_pad_remove)rVrurYr#r$rJ0s zHubertSamePadLayer.__init__cCs,|jdkr(|ddddd|j f}|S)Nr)rr\r#r#r$r^4s zHubertSamePadLayer.forwardr_r#r#rYr$r}/s r}cs0eZdZdZfddZddZddZZS)HubertFeatureEncoderz.Construct the features from raw audio waveformcstjdkr@tddgfddtjdD}n6jdkrdfddtjD}ntd jd t||_ d |_ d |_ dS) NgrouprrXcsg|]}t|ddqS)rr)rDr(irWr#r$r*Bsz1HubertFeatureEncoder.__init__..rlayercsg|]}t|dqS)r)rdrrr#r$r*Fsz`config.feat_extract_norm` is z), but has to be one of ['group', 'layer']FT) rIrJZfeat_extract_normrir6Znum_feat_extract_layersr.r ModuleList conv_layersgradient_checkpointing_requires_grad)rVrWrrYrr$rJ>s      zHubertFeatureEncoder.__init__cCs|D] }d|_qd|_dS)NF) parameters requires_gradrrVparamr#r#r$_freeze_parametersOs z'HubertFeatureEncoder._freeze_parameterscCsj|dddf}|jr"|jr"d|_|jD]<}|jr\|jr\|jr\dd}tjj|||}q(||}q(|S)NTcsfdd}|S)Ncs|Sr[r#inputsmoduler#r$custom_forward_szSHubertFeatureEncoder.forward..create_custom_forward..custom_forwardr#rrr#rr$create_custom_forward^s z;HubertFeatureEncoder.forward..create_custom_forward)rtrainingrrrtorchrv checkpoint)rV input_valuesr]Z conv_layerrr#r#r$r^Ts   zHubertFeatureEncoder.forward)r`rarb__doc__rJrr^rcr#r#rYr$r;s rcseZdZfddZZS)HubertFeatureExtractorcs8t|td|jjd|jjdjdtdS)Nz The class `zD` has been depreciated and will be removed in Transformers v5. Use `rz ` instead.)rIrJwarningswarnrZr` __bases__ FutureWarningrVrWrYr#r$rJos  zHubertFeatureExtractor.__init__)r`rarbrJrcr#r#rYr$rnsrcs$eZdZfddZddZZS)HubertFeatureProjectioncsXt|j|_|jr0tj|jd|jd|_t|jd|j |_ t |j |_ dS)Nr&Zeps)rIrJfeat_proj_layer_normrrerKlayer_norm_epsrfLinearrt projectionDropoutZfeat_proj_dropoutdropoutrrYr#r$rJzs  z HubertFeatureProjection.__init__cCs(|jr||}||}||}|Sr[)rrfrrr\r#r#r$r^s    zHubertFeatureProjection.forwardr_r#r#rYr$rys rc seZdZdZdeeeeedfdd Zej eedd d Z dej e ej e e ej e ej e ej ee ej e ej e e ej fd d dZ ZS)HubertAttentionz=Multi-headed attention from 'Attention Is All You Need' paperFT) embed_dim num_headsr is_decoderrHcst||_||_||_|||_|j||jkrNtd|jd|d|jd|_||_t j |||d|_ t j |||d|_ t j |||d|_ t j |||d|_dS)Nz;embed_dim must be divisible by num_heads (got `embed_dim`: z and `num_heads`: z).g)rH)rIrJrrrhead_dimr.scalingrrrk_projv_projq_projout_proj)rVrrrrrHrYr#r$rJs   zHubertAttention.__init__)tensorseq_lenbszcCs ||||j|jddSr~)viewrrrh contiguous)rVrrrr#r#r$_shapeszHubertAttention._shapeN)r]key_value_statespast_key_valuerlayer_head_maskoutput_attentionsrcCsx|dk }|\}} } |||j} |r\|dk r\|djd|jdkr\|d} |d} n|r|||d|} |||d|} n|dk r|||d|} |||d|} tj|d| gdd} tj|d| gdd} n(|||d|} |||d|} |j r | | f}||j d|j f}|| | |j |} | j |} | j |} | d}t| | dd}|||j | |fkrtd||j | |fd||dk r ||d| |fkrtd |d| |fd|| ||j | ||}| ||j | |}tjj|dd}|dk r||j fkrhtd |j fd|| dddd| ||j | |}| ||j | |}|r| ||j | |}| ||j | |}nd}tjj||j|jd }t|| }|||j | |j fkr4td ||j | |j fd|| ||j | |j }|dd}| || |j}||}|||fS) z#Input shape: Batch x Time x ChannelNrrlrr&rsz$Attention weights should be of size z , but is z!Attention mask should be of size z/Head mask for a single layer should be of size )prz `attn_output` should be of size )sizerrrrrrrcatrrrrr?Zbmmrhr.r functionalsoftmaxrrrr)rVr]rrrrrZis_cross_attentionrZtgt_lenr)Z query_statesZ key_statesZ value_statesZ proj_shapeZsrc_len attn_weightsZattn_weights_reshapedZ attn_probsZ attn_outputr#r#r$r^s~       "    zHubertAttention.forward)rFT)NNNNF)r`rarbrrfloatr8rJrTensorrrrr^rcr#r#rYr$rs4 rcs$eZdZfddZddZZS)HubertFeedForwardcsptt|j|_t|j|j|_ t |j t rDt |j |_n|j |_t|j|j|_t|j|_dSr[)rIrJrrZactivation_dropoutintermediate_dropoutrrtZintermediate_sizeintermediate_dense isinstanceZ hidden_actstrrintermediate_act_fn output_densehidden_dropoutoutput_dropoutrrYr#r$rJ(s  zHubertFeedForward.__init__cCs6||}||}||}||}||}|Sr[)rrrrrr\r#r#r$r^5s      zHubertFeedForward.forwardr_r#r#rYr$r's rcs&eZdZfddZdddZZS)HubertEncoderLayercsftt|j|j|jdd|_t|j |_ tj |j|j d|_ t||_tj |j|j d|_dS)NFrrrrr)rIrJrrtnum_attention_headsattention_dropout attentionrrrrrerrfr feed_forwardfinal_layer_normrrYr#r$rJAs  zHubertEncoderLayer.__init__NFcCsf|}|j|||d\}}}||}||}||}|||}||}|f}|rb||f7}|SNrr)rrrfrrrVr]rrZ attn_residualrr)outputsr#r#r$r^Ns     zHubertEncoderLayer.forward)NFr_r#r#rYr$r@s rcs,eZdZfddZejdddZZS)HubertAttnAdapterLayercsZt|j|_|j|_t|j|_t |j|j|_ t |_ t |j|j|_ dS)z Implements adapter modules directly with 3D tensor weight as parameters and without using ModuleList to speed up training throughput. N)rIrJadapter_attn_dimZ input_dimrtZ hidden_dimrrenormrlinear_1ZReLUact_fnlinear_2rrYr#r$rJds  zHubertAttnAdapterLayer.__init__)r]cCs,||}||}||}||}|Sr[)rrrrr\r#r#r$r^rs     zHubertAttnAdapterLayer.forward)r`rarbrJr FloatTensorr^rcr#r#rYr$rcs rcs8eZdZfddZdejeejedddZZ S) !HubertEncoderLayerStableLayerNormcstt|j|j|jdd|_t|j |_ tj |j|j d|_ t||_tj |j|j d|_t|dddk r~t||_nd|_dS)NFrrr)rIrJrrtrrrrrrrrerrfrrrgetattrr adapter_layerrrYr#r$rJ~s   z*HubertEncoderLayerStableLayerNorm.__init__NF)r]rrcCsz|}||}|j|||d\}}}||}||}||||}|jdk rb|||}|f}|rv||f7}|Sr)rfrrrrrrr#r#r$r^s      z)HubertEncoderLayerStableLayerNorm.forward)NF) r`rarbrJrrrr8r^rcr#r#rYr$r}s rcs<eZdZfddZd ejeejeeedddZ Z S) HubertEncodercsft|_t|_tjjjd|_ t j |_ t fddtjD|_d|_dS)Nrcsg|] }tqSr#)rr'rr#r$r*sz*HubertEncoder.__init__..FrIrJrWrkpos_conv_embedrrertrrfrrrrr6num_hidden_layerslayersrrrYrr$rJs   zHubertEncoder.__init__NFT)r]rroutput_hidden_states return_dictcs|rdnd}rdnd}|dk r|ddd|jd}d||<d|ddddddfj|jd}|t|jj}||jdd|jd|jd}| |} || }| |}| |}t } |j D]} |r||f}tg} |jr| |jjkrdnd } | r| r`|jrJ|jrJfd d }tjj|| ||}n| ||d }|d}| rjd }r||df}q|r||f}|stdd|||fDSt|||dS)Nr#r&rrlr?r+TFcsfdd}|S)Ncs|fSr[r#rrrr#r$rszLHubertEncoder.forward..create_custom_forward..custom_forwardr#rrrr$rsz4HubertEncoder.forward..create_custom_forwardrNNcss|]}|dk r|VqdSr[r#r(vr#r#r$ sz(HubertEncoder.forward..Zlast_hidden_stater] attentions) unsqueezerepeatrtor,rfinfominexpandrrfrr rr1rrW layerdroprrvrtupler rVr]rrrrZall_hidden_statesZall_self_attentionsZexpand_attention_maskZposition_embeddingsZdeepspeed_zero3_is_enabledrZdropout_probabilityZskip_the_layerrZ layer_outputsr#rr$r^sd   &         zHubertEncoder.forward)NFFT) r`rarbrJrrrrr8r^rcr#r#rYr$rs rcs&eZdZfddZdddZZS) HubertEncoderStableLayerNormcsft|_t|_tjjjd|_ t j |_ t fddtjD|_d|_dS)Nrcsg|] }tqSr#)rr'rr#r$r* sz9HubertEncoderStableLayerNorm.__init__..FrrrYrr$rJs  z%HubertEncoderStableLayerNorm.__init__NFTcs|rdnd}rdnd}|dk r|ddd|jd}d||<d|ddddddfj|jd}|t|jj}||jdd|jd|jd}| |} || }| |}t } |j D]} |r||f}t g} |jr| |jjkrdnd } | r| rR|jr<|jr<fd d }tjj|| ||}n| ||d }|d}| r\d }r||df}q||}|r||f}|stdd|||fDSt|||dS)Nr#r&rrlrrr+TFcsfdd}|S)Ncs|fSr[r#rrr#r$r:sz[HubertEncoderStableLayerNorm.forward..create_custom_forward..custom_forwardr#rrrr$r9szCHubertEncoderStableLayerNorm.forward..create_custom_forwardrrcss|]}|dk r|VqdSr[r#rr#r#r$rVsz7HubertEncoderStableLayerNorm.forward..r)rrrrr,rrrrrrr rr1rrWrrrvrrfrr rr#rr$r^sd   &         z$HubertEncoderStableLayerNorm.forward)NFFTr_r#r#rYr$rs  rc@s\eZdZdZeZdZdZdZddZ ddd Z e e j efd d d Zee j d ddZdS)HubertPreTrainedModelz An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. hubertrTc Cs&t|tjr$|jjjd|jjdnt|tjtj frR|j j |jj dnt|tj rtrddl}t|drt|dr|jj|j|jgddtj|jjW5QRXq|jj|jddtj|jjW5QRXntj|jjt|tjtj fr"|j dk r"|j j dS) zInitialize the weightsr)meanZstdrrNr{r|rp)rrrrqdataZnormal_rWZinitializer_rangererjrHZzero_Zfill_rNr rxrwryrzr{r|initZkaiming_normal_)rVrrxr#r#r$ _init_weightsis     z#HubertPreTrainedModel._init_weightsFcCst|ttfr||_dSr[)rrrr)rVrvaluer#r#r$_set_gradient_checkpointingsz1HubertPreTrainedModel._set_gradient_checkpointing)rAcCs4dd}t|jj|jjD]\}}||||}q|S)zH Computes the output length of the convolutional layers cSstj|||dddS)Nfloor)Z rounding_moder)rdiv)rrFrGr#r#r$_conv_out_lengthszPHubertPreTrainedModel._get_feat_extract_output_lengths.._conv_out_length)ziprWrOrP)rVrAr rFrGr#r#r$ _get_feat_extract_output_lengthssz6HubertPreTrainedModel._get_feat_extract_output_lengths)feature_vector_lengthrcCs~||dtj}|jd}tj||f|j|jd}d|tj |jd|jd|df<| dg d dg }|S)Nr&r)r,devicer)r) r r3rrlongrr7r,rr:flipZcumsumr8)rVr rZoutput_lengthsr@r#r#r$"_get_feature_vector_attention_masks "z8HubertPreTrainedModel._get_feature_vector_attention_maskN)F)r`rarbrr config_classZbase_model_prefixZmain_input_nameZsupports_gradient_checkpointingrrrr LongTensorrr rr#r#r#r$r^s ra! Hubert was proposed in [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) by Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed. This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving etc.). This model is a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config ([`HubertConfig`]): 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. a Args: 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]` or a `numpy.ndarray`, *e.g.* via 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 [`Wav2Vec2Processor.__call__`] for details. attention_mask (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing convolution and attention on padding token indices. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) `attention_mask` should only be passed if the corresponding processor has `config.return_attention_mask == True`. For all models whose processor has `config.return_attention_mask == False`, such as [hubert-base](https://huggingface.co/facebook/hubert-base-ls960), `attention_mask` should **not** be passed to avoid degraded performance when doing batched inference. For such models `input_values` should simply be padded with 0 and passed without `attention_mask`. Be aware that these models also yield slightly different results depending on whether `input_values` is padded or not. output_attentions (`bool`, *optional*): Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned tensors for more detail. output_hidden_states (`bool`, *optional*): Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for more detail. return_dict (`bool`, *optional*): Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. z`The bare Hubert Model transformer outputting raw hidden-states without any specific head on top.c seZdZedfdd Zd ejeejeejdddZ e e e e edd eejeejeejeeeeeeeee fd d d ZZS) HubertModelrcszt|||_t||_t||_|jdks:|jdkrRt t |j |_|jrdt||_n t||_|dS)Nr)rIrJrWrfeature_extractorrfeature_projectionmask_time_probmask_feature_probr ParameterrrrtZuniform_masked_spec_embedZdo_stable_layer_normrencoderr post_initrrYr#r$rJs     zHubertModel.__init__N)r]mask_time_indicesrcCst|jdds|S|\}}}|dk r<|j|j||<nZ|jjdkr|jrt||f|jj|jj ||jj d}t j ||j t jd}|j|j||<|jjdkr|jrt||f|jj|jj|jjd}t j ||j t jd}|dddfd|d}d||<|S) z Masks extracted features along time axis and/or along feature axis according to [SpecAugment](https://arxiv.org/abs/1904.08779). Zapply_spec_augmentTNr)rrrr)rr,)rrrr&)rrWrrrr,rrrCZmask_time_lengthZmask_time_min_masksrrrr8rZmask_feature_lengthZmask_feature_min_masksr)rVr]rrr@r"rtZmask_feature_indicesr#r#r$_mask_hidden_statess4 zHubertModel._mask_hidden_states) output_typer)rrrrrrrc Cs|dk r |n|jj}|dk r |n|jj}|dk r4|n|jj}||}|dd}|dk rl||jd|}||}|j ||d}|j |||||d} | d}|s|f| ddSt || j | j dS)aZ Returns: Example: ```python >>> from transformers import AutoProcessor, HubertModel >>> from datasets import load_dataset >>> import soundfile as sf >>> processor = AutoProcessor.from_pretrained("facebook/hubert-large-ls960-ft") >>> model = HubertModel.from_pretrained("facebook/hubert-large-ls960-ft") >>> def map_to_array(batch): ... speech, _ = sf.read(batch["file"]) ... batch["speech"] = speech ... return batch >>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation") >>> ds = ds.map(map_to_array) >>> input_values = processor(ds["speech"][0], return_tensors="pt").input_values # Batch size 1 >>> hidden_states = model(input_values).last_hidden_state ```Nrrl)rrrrrrr)rWrruse_return_dictrrhrrrrrr r]r) rVrrrrrrZextract_featuresr]Zencoder_outputsr#r#r$r^ s2&   zHubertModel.forward)NN)NNNNN)r`rarbrrJrrrrrrHUBERT_INPUTS_DOCSTRINGrr _CONFIG_FOR_DOCrr8rrr^rcr#r#rYr$rs0 .  rzdHubert Model with a `language modeling` head on top for Connectionist Temporal Classification (CTC).c seZdZdeedfdd ZddZddZd d Zd d Z e e e e eeeed deejeejeeeeeeeejeeefdddZZS) HubertForCTCN) target_langcs~t|t||_t|j|_||_|j dkrFt d|j dt |dr\|j r\|jn|j}t||j |_|dS)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: `HubertForCTC.from_pretrained(..., vocab_size=vocab_size)`. or define `vocab_size` of your model's configuration. add_adapter)rIrJrrrrZ final_dropoutrr% vocab_sizer.rZrwr&output_hidden_sizertrlm_headr)rVrWr%r(rYr#r$rJps    zHubertForCTC.__init__cCsr|j}|dk r2t|jdddkr2td|dn<|dkrXt|jdddk rXtdn|dk rn|j|dddS)a' This method overwrites [`~PreTrainedModel.tie_weights`] so that adapter weights can be correctly loaded when passing `target_lang=...` to `from_pretrained(...)`. This method is **not** supposed to be called by the user and is prone to be changed in the future. NrzCannot pass `target_lang`: z- if `config.adapter_attn_dim` is not defined.z)By default `target_lang` is set to 'eng'.T)Z force_load)r%rrWr.loggerinfoZ load_adapter)rVr%r#r#r$ tie_weightss  zHubertForCTC.tie_weightscCstdt|dS) Calling this function will disable the gradient computation for the feature encoder so that its parameter will not be updated during training. The method `freeze_feature_extractor` is deprecated and will be removed in Transformers v5.Please use the equivalent `freeze_feature_encoder` method instead.Nrrrfreeze_feature_encoderrVr#r#r$freeze_feature_extractors z%HubertForCTC.freeze_feature_extractorcCs|jjdSr-Nrrrr1r#r#r$r0sz#HubertForCTC.freeze_feature_encodercCs|jD] }d|_q dSz 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. FNrrrrr#r#r$freeze_base_modelszHubertForCTC.freeze_base_model)rrrexpected_output expected_lossrrrrrlabelsrc Csf|dk r |n|jj}|j|||||d}|d}||}||} d} |dk r"||jjkrttd|jj|dk r|ntj |tj d}| | d tj } |dk} | d} || }tjj| dtjddd}tjjjd d ,tjj||| | |jj|jj|jjd } W5QRX|sR| f|td}| dk rN| f|S|St| | |j|jd S) a labels (`torch.LongTensor` of shape `(batch_size, target_length)`, *optional*): Labels for connectionist temporal classification. Note that `target_length` has to be smaller or equal to the sequence length of the output logits. Indices are selected in `[-100, 0, ..., config.vocab_size - 1]`. All labels set to `-100` are ignored (masked), the loss is only computed for labels in `[0, ..., config.vocab_size - 1]`. Nr rz$Label values must be <= vocab_size: r+r&)rsr,rF)enabled)blankZ reductionZ zero_infinitylosslogitsr]r)rWr!rrr)rr'r.rZ ones_likerr r3rZ masked_selectrrZ log_softmaxZfloat32rhbackendsZcudnnflagsZctc_lossZ pad_token_idZctc_loss_reductionZctc_zero_infinity_HIDDEN_STATES_START_POSITIONr r]r)rVrrrrrr;rr]r@r?rAZ labels_maskZtarget_lengthsZflattened_targetsZ log_probsoutputr#r#r$r^sR      zHubertForCTC.forward)N)NNNNN)r`rarbrrrJr,r2r0r7rr"r_CHECKPOINT_FOR_DOCr r#_CTC_EXPECTED_OUTPUT_CTC_EXPECTED_LOSSrrr8rrr^rcr#r#rYr$r$js6   r$z Hubert Model with a sequence classification head on top (a linear layer over the pooled output) for tasks like SUPERB Keyword Spotting. c seZdZfddZddZddZddZeee e e e d e ed deejeejeeeeeeeejeee fd d dZZS)HubertForSequenceClassificationcst|t|dr$|jr$tdt||_|jd}|jrTt t |||_ t |j|j|_t |j|j|_|dS)Nr&z]Sequence classification does not support the use of Hubert adapters (config.add_adapter=True)r)rIrJrwr&r.rrruse_weighted_layer_sumrrrr< layer_weightsrrtZclassifier_proj_size projector num_labels classifierr)rVrWZ num_layersrYr#r$rJs   z(HubertForSequenceClassification.__init__cCstdt|dS)z Calling this function will disable the gradient computation for the feature encoder so that its parameters will not be updated during training. r.Nr/r1r#r#r$r2!s z8HubertForSequenceClassification.freeze_feature_extractorcCs|jjdSr3r4r1r#r#r$r0-sz6HubertForSequenceClassification.freeze_feature_encodercCs|jD] }d|_q dSr5r6rr#r#r$r74sz1HubertForSequenceClassification.freeze_base_modelZaudio)rrrZmodalityr8r9Nr:cCsf|dk r |n|jj}|jjr dn|}|j|||||d}|jjr|t}tj|dd}tjj |j dd} || dddj dd}n|d}| |}|dkr|jdd} n<||jd|} d|| <|j dd| j dd dd} || } d} |dk r"t}|| d|jj| d} |sR| f|td}| dk rN| f|S|St| | |j|jd S) a 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). NTr rrr&rrr>)rWr!rIrrCrstackrrrrJrr3rKrrrrMrrLr r]r)rVrrrrrr;rr]Z norm_weightsZ pooled_outputZ padding_maskr@r?Zloss_fctrDr#r#r$r^<sF     z'HubertForSequenceClassification.forward)NNNNN)r`rarbrJr2r0r7rr"r_SEQ_CLASS_CHECKPOINTr r#_SEQ_CLASS_EXPECTED_OUTPUT_SEQ_CLASS_EXPECTED_LOSSrrrr8rrr^rcr#r#rYr$rHs6    rH)Nr)Grrtypingrrrnumpyr/rZtorch.utils.checkpointrZtorch.nnrZ activationsrZintegrations.deepspeedr Zmodeling_outputsr r r Zmodeling_utilsr rvrrrrrZconfiguration_hubertrZ get_loggerr`r*rCr#rEZ_EXPECTED_OUTPUT_SHAPErFrGrOrPrQZ$HUBERT_PRETRAINED_MODEL_ARCHIVE_LISTrrrZndarrayrCModulerDrdrirkr}rrrrrrrrrrrZHUBERT_START_DOCSTRINGr"rr$rHr#r#r#r$s           x( 3 #.X[D&