o wi@sddlmZddlmZmZmZddlZddlmZddl m Z ddl m Z m Z ddlmZddlmZdd lmZdd lmZdd lmZdd lmZmZdd lmZmZddlmZm Z ddl!m"Z"ddl#m$Z$m%Z%m&Z&m'Z'm(Z(ddl)m*Z*ddl+m,Z,m-Z-ddl.m/Z/e(0e1Z2ee&ddGddde%Z3e&dde&Gddde Z4edGdddej5Z6Gd d!d!ej5Z7Gd"d#d#ej5Z8d$d%Z9dLd&d'Z:d(ej;d)ed4e>fd5d6Z?Gd7d8d8ej5Z@Gd9d:d:eZAe&Gd;d<dd>ej5ZCGd?d@d@ee$ZDe&dAdGdBdCdCe4eZEGdDdEdEej5ZFe&GdFdGdGe4ZGe&dHdGdIdJdJe4e/ZHgdKZIdS)N) dataclass)CallableOptionalUnionN)ACT2FN)Cache DynamicCache)GenerationMixin)use_kernel_forward_from_hub)create_causal_mask)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack) LossKwargs ModelOutputauto_docstringcan_return_tuplelogging) AutoModel) CsmConfigCsmDepthDecoderConfig)CsmGenerationMixinz: Base class for the model autoregressive outputs. ) custom_introc@seZdZUdZdZeejed<dZ ejed<dZ ee e ejed<dZ ee ejdfed<dZ ee ejdfed<dZeejed <dZejed <dZee e ejed <dZee ejdfed <dZee ejdfed <dZeejed<dS)CsmOutputWithPasta# loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction). logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`): Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`) Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. depth_decoder_loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction) of the depth decoder model. depth_decoder_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`): Prediction scores of the depth decoder (scores for each vocabulary token before SoftMax). depth_decoder_past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`) depth_decoder_hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`): Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states of the model at the output of each layer plus the optional initial embedding outputs. depth_decoder_attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`): Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length, sequence_length)`. backbone_loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction) of the backbone model. Nlosslogitspast_key_values. hidden_states attentionsdepth_decoder_lossdepth_decoder_logitsdepth_decoder_past_key_valuesdepth_decoder_hidden_statesdepth_decoder_attentions backbone_loss)__name__ __module__ __qualname____doc__r#rtorch FloatTensor__annotations__r$r%tupler&r'r(r)r*r+r,r-r6r6a/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/transformers/models/csm/modeling_csm.pyr"0s r"z[ The bare Csm Model outputting raw hidden-states without any specific head on top. c@sDeZdZeZdZdZdgZdgZdZ dZ dZ dZ dZ dZddZdS)CsmPreTrainedModelmodelTCsmDecoderLayerr%cCs|jj}t|tjr"|jjjd|d|jdur |jj dSdSt|tj rC|jjjd|d|j durA|jj|j  dSdSt|t ra|j }t|dD] }|jj|jd|dqQdSt|tro|jjddSdS)N)meanstdrg?)configinitializer_range isinstancennLinearweightdatanormal_biaszero_ Embedding padding_idxCsmCodebooksHead num_codebooksrange CsmRMSNormfill_)selfmoduler=rKir6r6r7 _init_weightsws&      z CsmPreTrainedModel._init_weightsN)r.r/r0r config_classbase_model_prefixsupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_2_supports_sdpa_supports_cache_class_supports_quantized_cache_supports_static_cache_supports_attention_backendrRr6r6r6r7r8bs r8RMSNormcs.eZdZdfdd ZddZddZZS) rMư>cs&ttt||_||_dS)z9 CsmRMSNorm is equivalent to T5LayerNorm N)super__init__rA Parameterr2onesrCvariance_epsilon)rO hidden_sizeeps __class__r6r7ras  zCsmRMSNorm.__init__cCsJ|j}|tj}|djddd}|t||j}|j||S)NrT)keepdim) dtypetor2float32powr<rsqrtrdrC)rOr& input_dtypevariancer6r6r7forwards  zCsmRMSNorm.forwardcCst|jjd|jS)Nz, eps=)r5rCshaperdrOr6r6r7 extra_reprszCsmRMSNorm.extra_repr)r_)r.r/r0rarrru __classcell__r6r6rgr7rMsrMcs8eZdZddeffdd ZeeddZZ S)CsmRotaryEmbeddingNr>cstt|dr|jdur|jd|jd|_nd|_|j|_|j|_||_ t |j|_ | |j |\}|_ |j d|dd|j|_dS)N rope_scaling rope_typetypedefaultinv_freqF persistent)r`rahasattrrxgetrymax_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenr>r rope_init_fnattention_scalingregister_bufferr|original_inv_freq)rOr>devicer|rgr6r7ras   zCsmRotaryEmbedding.__init__c Cs|jddddf|jddd|j}|dddddf}t|jjtr6|jjdkr6|jjnd}t j |dd+|| dd}t j ||fdd }| |j}||j} Wdn1smwY|j|jd | j|jd fS) NrrirmpscpuF) device_typeenabledrdim)rk)r|floatexpandrsrlrr@rzstrr2autocast transposecatcosrsinrk) rOx position_idsinv_freq_expandedposition_ids_expandedrfreqsembrrr6r6r7rrs0&zCsmRotaryEmbedding.forwardN) r.r/r0rrar2no_gradrrrrvr6r6rgr7rws rwc$eZdZfddZddZZS)CsmMLPcsxt||_|j|_|j|_tj|j|j|jd|_tj|j|j|jd|_ tj|j|j|jd|_ t |j |_ dS)NrF)r`rar>reintermediate_sizerArBmlp_bias gate_projup_proj down_projr hidden_actact_fnrOr>rgr6r7ras zCsmMLP.__init__cCs$||||||}|Sr)rrrr)rOrrr6r6r7rrs zCsmMLP.forwardr.r/r0rarrrvr6r6rgr7rs  rcCsH|dd|jddf}|d|jdddf}tj| |fddS)z*Rotates half the hidden dims of the input..Nrirr)rsr2r)rx1x2r6r6r7 rotate_halfsrcCsD||}||}||t||}||t||}||fS)aApplies Rotary Position Embedding to the query and key tensors. Args: q (`torch.Tensor`): The query tensor. k (`torch.Tensor`): The key tensor. cos (`torch.Tensor`): The cosine part of the rotary embedding. sin (`torch.Tensor`): The sine part of the rotary embedding. position_ids (`torch.Tensor`, *optional*): Deprecated and unused. unsqueeze_dim (`int`, *optional*, defaults to 1): The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2. Returns: `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding. ) unsqueezer)qkrrr unsqueeze_dimq_embedk_embedr6r6r7apply_rotary_pos_embs  rr&n_repreturncCs^|j\}}}}|dkr |S|dddddddddf|||||}||||||S)z This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim) rN)rsrreshape)r&rbatchnum_key_value_headsslenhead_dimr6r6r7 repeat_kvs 0rr;rPquerykeyvalueattention_maskscalingdropoutc Kst||j}t||j} t||dd|} |dur3|ddddddd|jdf} | | } tjj| dtj d |j } tjj | ||j d} t| | } | dd} | | fS)Nrrri)rrk)ptrainingr)rnum_key_value_groupsr2matmulrrsrA functionalsoftmaxrmrlrkrr contiguous) rPrrrrrrkwargs key_states value_states attn_weights causal_mask attn_outputr6r6r7eager_attention_forwards & rcseZdZdZdedeffdd Z  ddejde ejejfd e ejd e e d e ej d e ed e eje eje e ejffddZZS) CsmAttentionz=Multi-headed attention from 'Attention Is All You Need' paperr> layer_idxcst||_||_t|d|j|j|_|j|j|_ |jd|_ |j |_ d|_ t j|j|j|j|jd|_t j|j|j|j|jd|_t j|j|j|j|jd|_t j|j|j|j|jd|_dS)NrgTr)r`rar>rgetattrrenum_attention_headsrrrrattention_dropout is_causalrArBattention_biasq_projk_projv_projo_projrOr>rrgr6r7ras(   zCsmAttention.__init__Nr&position_embeddingsrpast_key_valuecache_positionrrcKs$|jdd}g|d|jR}|||dd} |||dd} |||dd} |\} } t| | | | \} } |durW| | |d}|| | |j |\} } t }|j j dkret |j j }||| | | |f|jsqdn|j|jd|\}}|jg|dR}||}||fS)Nrirr)rrreagerr;)rr)rsrrviewrrrrupdaterrr>_attn_implementationrrrrrrr)rOr&rrrrr input_shape hidden_shape query_statesrrrr cache_kwargsattention_interfacerrr6r6r7rr2s8      zCsmAttention.forward)NN)r.r/r0r1rintrar2Tensorr5rr LongTensorrr rrrvr6r6rgr7rs(rcseZdZdedeffdd Z       ddejdeejd eej d ee d ee d ee d eej dee ejejfde ede ejee ejejfffddZZS)r:r>rcsRt|j|_t||d|_t||_t|j|jd|_ t|j|jd|_ dS)N)r>rrf) r`rarer self_attnrmlprM rms_norm_epsinput_layernormpost_attention_layernormrrgr6r7ra_s  zCsmDecoderLayer.__init__NFr&rrroutput_attentions use_cacherrrrc Kst|} ||}|jd||||||||d| \}} | |}|} ||}||}| |}|f} |r8| | f7} | S)N)r&rrrrrrrr6)rrrr) rOr&rrrrrrrrresidualself_attn_weightsoutputsr6r6r7rris.      zCsmDecoderLayer.forward)NNNFFNN)r.r/r0rrrar2rrrrboolr5rr r3rrrvr6r6rgr7r:^s<    r:cseZdZeZfddZddZddZee          dde j d e e j d e e jd e e j d e ed e e j de ede ede ede e j deedeeeffddZZS)CsmDepthDecoderModelcstj|_j|_tjjj|_ t fddt j D|_ tjjd|_td|_d|_tjjjdd|_|dS)Ncg|]}t|qSr6r:.0rr>r6r7 z1CsmDepthDecoderModel.__init__..rrFr)r`ra pad_token_idrI vocab_sizerArHrKbackbone_hidden_size embed_tokens ModuleListrLnum_hidden_layerslayersrMrernormrw rotary_embgradient_checkpointingrBinputs_embeds_projector post_initrrgrr7ras   zCsmDepthDecoderModel.__init__cC|jSrrrtr6r6r7get_input_embeddingsz)CsmDepthDecoderModel.get_input_embeddingscC ||_dSrrrOrr6r6r7set_input_embeddings z)CsmDepthDecoderModel.set_input_embeddingsN input_idsbackbone_last_hidden_staterrr% inputs_embedsrroutput_hidden_statesrflash_attn_kwargsrc  Ks\|durtjstdd}|dur|n|jj}| dur | n|jj} |dur*|n|jj}|du|duAr:t d|j rI|j rI|rItdd}|rR|durRt }| dur|dur^| nd} |duri|jdn|jd} |duru|jn|j}tj| | | |d} |durtj| ddd }||j}|||}| ddk}|dur||dddf<n tjs|rtd ||}t|j||| ||d }|}| d}|||}| rd nd}|rd nd}|jd|jjD]'}| r||f7}||f|||||| |d | }|d}|r||df7}q||}| r!||f7}t||r(|nd||dS)aJ backbone_last_hidden_state (`torch.FloatTensor` of shape `(batch_size, backbone_hidden_size)`, *optional*): The last hidden state of the backbone model. Such input is required when the first codebook token (the one generated by the backbone model) is provided in the `input_ids` argument. NzCustom `position_ids` were provided but will be ignored. CSM depth decoder automatically determines position_ids from `cache_position` and as it requires them to be identical across the batch, the provided position_ids will be ignored.z;You must specify exactly one of input_ids or inputs_embeds.X`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.Frrr)minzvWhen the first codebook token is provided, `backbone_last_hidden_state` should also be provided for correct inference.r> input_embedsrrr%rr6rrrrrrrlast_hidden_stater%r&r')r2compiler is_compilinglogger warning_oncer>rrr ValueErrorr rr get_seq_lengthrsrarangeclamprrwarningr r rr rrrr)rOrrrrr%rrrrrrpast_seen_tokensinputs_seq_lengthr codebook_idxsoffsetinput_ids_are_first_codebookrr&rall_hidden_statesall_self_attns decoder_layer layer_outputsr6r6r7rrs              zCsmDepthDecoderModel.forward) NNNNNNNNNN)r.r/r0rrSrarrrrr2rrr3rrrrr rr5rrrrvr6r6rgr7rsT      rcs&eZdZfddZdddZZS)rJcs0t||_tt|jd|||_dSNr)r`rarKrArbr2emptyrC)rOrerKrrgr6r7ra(s  zCsmCodebooksHead.__init__Ncsf|durjd}|jt|n |d}|j|fddtjdDtjddS)Nrc s2g|]}tjdd|ddf|jqSr)rArlinearT)r codebook_idxcodebook_weightr&r6r7r5s$z,CsmCodebooksHead.forward..rr)rsrCr2r(rLstack)rOr&r seq_lengthr-r6r9r7rr-s    zCsmCodebooksHead.forwardrrr6r6rgr7rJ's rJc@s eZdZdS)KwargsForCausalLMN)r.r/r0r6r6r6r7r=>s r=a$ The CsmDepthDecoder Model transformer, with a [`CsmCodebooksHead`] on top, which can be seen a position-specific language modeling head, allowing to use a different linear layer for each codebook (e.g. position 0 is the first codebook and uses the first codebook head, etc.) c!sJeZdZdZdZdZfddZddZddZdd Z d d Z e e            dd e jdee jdee jdee jdeeeee jfdee jdee jdeedeedeedee jdeee jfdeedeeeffddZ    d d e jdeedee jdee jdee jf fdd ZZS)!CsmDepthDecoderForCausalLMNcs>t|t||_|j|_t|j|j|j|_| dSr) r`rarr9rrJrerKcodebooks_headr rrgr6r7raMs   z#CsmDepthDecoderForCausalLM.__init__cC|jjSrr9rrtr6r6r7rVz/CsmDepthDecoderForCausalLM.get_input_embeddingscC ||j_dSrrArr6r6r7rY z/CsmDepthDecoderForCausalLM.set_input_embeddingscCrrr9)rOdecoderr6r6r7 set_decoder\rz&CsmDepthDecoderForCausalLM.set_decodercCr rrErtr6r6r7 get_decoder_rz&CsmDepthDecoderForCausalLM.get_decoderrrrrrr%rlabelsrrrrlogits_to_keeprrc  Ks| dur| n|jj} | dur| n|jj} |jd|||||||| | | d | }|d}t| trA| dkr:tdd}n t| d}n| }||dd|ddf| durW| |nd}|}d}|dur}|dddf}|j d|d|jj |d| }t |||j |j |jdS) a backbone_last_hidden_state (`torch.FloatTensor` of shape `(batch_size, backbone_hidden_size)`, *optional*): The last hidden state of the backbone model. Such input is required when the first codebook token (the one generated by the backbone model) is provided in the `input_ids` argument. labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. N) rrrrr%rrrrrrr.)r$rIr shift_labels)r#r$r%r&r'r6)r>rrr9r@rslicer?r loss_functionrrr%r&r')rOrrrrr%rrIrrrrrJrrr& slice_indicesr$r#rKr6r6r7rrbsT   & z"CsmDepthDecoderForCausalLM.forwardc  sHtj|||||fi|}|dddk}|s|d|d|S)Nrrrr)r`prepare_inputs_for_generationpop) rOrr%rrrr model_inputsis_first_generation_steprgr6r7rOs   z8CsmDepthDecoderForCausalLM.prepare_inputs_for_generation) NNNNNNNNNNNrNNNN)r.r/r0_tied_weights_keys_tp_plan_pp_planrarrrGrHrrr2rrr3rrrlistrrrr=r5rrrrOrvr6r6rgr7r>As        Or>cr)CsmBackboneModelEmbeddingscsDtt|j|j|j|_|jdt |j|jdddS)Naudio_tokens_offsetsFr}) r`rarArHrKrreembed_audio_tokensrr2r(rrgr6r7ras  z#CsmBackboneModelEmbeddings.__init__cCs |||j}|jdd}|S)Nrr)rZrYsum)rOrrr6r6r7rrs z"CsmBackboneModelEmbeddings.forwardrr6r6rgr7rXs rXcseZdZfddZddZddZee         ddee j d ee j d ee j d ee d ee j d eedeedeedee j deedefddZZS)CsmBackboneModelcsvtj|_j|_t|_tfddt j D|_ t j jd|_td|_d|_|dS)Ncrr6rrrr6r7rrz-CsmBackboneModel.__init__..rrF)r`rarrIrrXrrArrLrrrMrerrrwr r r rrgrr7ras    zCsmBackboneModel.__init__cCr rrrtr6r6r7rrz%CsmBackboneModel.get_input_embeddingscCrrrrr6r6r7rrz%CsmBackboneModel.set_input_embeddingsNrrrr%rrrrrrrc  Ks|dur|n|jj}|dur|n|jj}|dur|n|jj}|du|duAr*td|jr9|jr9|r9tdd}t |t dt fsFtd|durO| |}|rX|durXt }| durt|durd|nd} tj| | |jd|jd} |dur}| d}t|j||| ||d } |} || |}|rd nd}|rd nd}|jd|jjD]&}|r|| f7}|| f| ||||| |d | }|d} |r||df7}q|| } |r|| f7}t| |r|nd||d S) a& input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length, num_codebooks) or (batch_size, sequence_length)`): 1. (batch_size, sequence_length): corresponds to the input sequence prepared with the processor from the text prompt. Such input requires `input_values` to be provided so that audio can be encoded in codebook tokens and then merged with the text tokens. 2. (batch_size, sequence_length, num_codebooks): codebook tokens generated during the autoregressive decoding. Such input is not meant to be used by end users. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) Nz:You must specify exactly one of input_ids or inputs_embedsrFzBThe `past_key_values` should be either a `Cache` object or `None`.rrrrr6rr )r>rrrr&r rr$r%r@rzrrr r'r2r(rsrrr r rrrr)rOrrrr%rrrrrrr+rr&rr0r1r2r3r6r6r7rrs            zCsmBackboneModel.forward) NNNNNNNNN)r.r/r0rarrrrrr2rrrr3rrr rrrrvr6r6rgr7r\sL     r\z The Csm model consists of two llama-like auto-regressive transformer models: a backbone model that predicts the first codebook token and a depth decoder that predicts the other codebook tokens. c#seZdZddgZfddZddZddZd d Zd d Zd dZ e fddZ fddZ    d*de ejde ejde ejde ejde ejf ddZ    d*dejde ede ejde ejde ejf fdd Zee             !d+dejde ejde ejde ejd"e ejde eeeejfde ejde ejd#e ed$e ed%e ede ejd&eeejfd'eedeeeffd(d)ZZ S),CsmForConditionalGenerationz5backbone_model.embed_tokens.embed_audio_tokens.weightz'depth_decoder.model.embed_tokens.weightcspt||j|_tj|j|jdd|_t|j|j|_ t ||_ t |j|_t|j|_|dS)NFr)r`rarrArBrelm_headrHtext_vocab_sizeembed_text_tokensr\ _from_configbackbone_modelr>depth_decoder_config depth_decoderr from_config codec_config codec_modelr rrgr6r7rais   z$CsmForConditionalGeneration.__init__cCr@rrbrrtr6r6r7rurBz0CsmForConditionalGeneration.get_input_embeddingscCrCrrhrr6r6r7rxrDz0CsmForConditionalGeneration.set_input_embeddingscCr rr^rtr6r6r7get_output_embeddings{rz1CsmForConditionalGeneration.get_output_embeddingscCrrri)rOnew_embeddingsr6r6r7set_output_embeddings~rz1CsmForConditionalGeneration.set_output_embeddingscCs(|jjr||jjj|jjjdSdSr)r>tie_codebooks_embeddings_tie_or_clone_weightsrbrrZrdr9rtr6r6r7 _tie_weightss z(CsmForConditionalGeneration._tie_weightscs|ddrtj|i|\}}n tj|i|}dtfddt|jD}t|jjddi||D] }t |j|q?d|vrR||fS|S)Noutput_loading_infoFdepth_decoder_cs(i|]\}}|r|d|qSr) startswith)rattrrprefix prefix_lenr6r7 s  z?CsmForConditionalGeneration.from_pretrained.._from_model_config) rr`from_pretrainedlenvarsgeneration_configitemsrdrdelattr)clsargsrr9 loading_infodepth_decoder_attrsrsrgrtr7rys   z+CsmForConditionalGeneration.from_pretrainedcsVd}|jj}|dd|D] \}}t|j|||qtj|i|dS)Nrqtransformers_version)rdr| to_diff_dictrPr}setattrr`save_pretrained)rOrrrurrsrrgr6r7rs   z+CsmForConditionalGeneration.save_pretrainedNr input_valuesinput_values_cutoffsrIrcsF||}|durtj|d}||dk}||dk}tj||jd t |d}|| dk}t jg}t ||D]?\} } | | dk} t| jddD]+} | | } | | d} | d| | f}|j| d}|jdd}||dqUqBtdd |Dtfd d |D}|j|}Wdn1swY|jj}||k}|j|}||||<tjdd|jjf|jtjd |jj}|j|d}||jj k}|!|"d||<|dur| d!dd|jj}||||<|||<|d kj#dd}d||d|dddf<|}||dS)a Merges the input_ids and input_values to produce a single inputs_embeds tensor: 1 - Infers the codec model on the input_values to retreive codebook token. 2 - Embeds codebook tokens and places them at the correct positions in the inputs_embeds tensor. 3 - If labels are provided, expands them to match codebook dimensions and position the target codebook tokens in the inputs_embeds tensor. Args: input_ids (`torch.Tensor` of shape `(batch_size, sequence_length)`): The input ids to embed. input_values (`torch.Tensor` of shape `(batch_size, channels, audio_sequence_length)`): The audio input values to embed. input_values_cutoffs (`torch.Tensor` of shape `(batch_size, max_num_audio)`): The cutoffs of the audio input values relative to its batch index, padded with -1 when no audio. Nrrrrrir.css|]}|jdVqdS)rN)rsrelr6r6r7 szQCsmForConditionalGeneration._merge_input_ids_with_input_values..c s,g|]}tj|ddd|jdfqS)r)rArpadrsrmax_audio_framesr6r7rs,zRCsmForConditionalGeneration._merge_input_ids_with_input_values..)rrkiTas_tuple)rrI)$r`rArrdiffr2r(maxrrrzrrziprLrsrgencode audio_codesrappendr;get_audio_codes_maskr>audio_token_idrbrrcrKlongcodebook_eos_token_idsqueezeaudio_eos_token_idrepeatr[nonzero)rOrrrrIr audio_lengthsinput_values_maskaudio_tokens_listbatch_input_valuesbatch_input_values_cutoffsrQ start_idxend_idx audio_batch codec_outputs codebook_idsbatched_audio_token_idsaudio_codes_maskraudio_token_mask audio_embedsaudio_eos_frame_idsaudio_eos_embedsaudio_eos_token_masklabels_expanded depth_decoder_ignore_frames_idxsr6rr7"_merge_input_ids_with_input_valuess\            z>CsmForConditionalGeneration._merge_input_ids_with_input_valuesr%rrrc  stjd |||||d|}|dur>|jdkr>|ddur>|j||d|d|dd}||d|ddd|S) N)rr%rrrrrrrrI)rrrrI)rrIrr6)r`rOndimrrr) rOrr%rrrrrQ merged_inputsrgr6r7rOs(  z9CsmForConditionalGeneration.prepare_inputs_for_generationrrrrrrJrc Ks| dur| n|jj} | dur| n|jj} |dur/|jdkr/|||||}|d}|d}d}|jd|||||| | | | d |}|d}t| trPt| dn| }| |dd|ddf}d}d}d}d}|dur|dddddf}|j d|||jj d|}|ddddddfd kj d d }||d d|jj df}tjj|d dd}|jdd}||d|ddddf}||}|j||| | | d|d}|j}||}t|||||j|j|j|dur|jnd|dur|jnd|dur|jnd|dur |jd Sdd S)a input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length, num_codebooks) or (batch_size, sequence_length)`): 1. (batch_size, sequence_length): corresponds to the input sequence prepared with the processor from the text prompt. Such input requires `input_values` to be provided so that audio can be encoded in codebook tokens and then merged with the text tokens. 2. (batch_size, sequence_length, num_codebooks): codebook tokens generated during the autoregressive decoding. Such input is not meant to be used by end users. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) input_values_cutoffs (`torch.Tensor` of shape `(batch_size, max_num_audio)`, *optional*): Specify the end positions of audio segments within each batch entry, relative to the concatenated audio input. If a batch entry has fewer segments than the maximum, it is padded with -1. For example, in a batch of 2 sequences where the first contains 2 audio segments of length l1, and the second contains 1 audio segment of length l2, the input_values_cutoffs would be: [[l1, 2 * l1], [l2, -1]]. labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked language modeling loss. Indices should be in `[config.audio_token_id, -100, -101]`. Requires targeted `input_values` to be provided as audio tokens will be infered from it using the `codec_model`. - `config.audio_token_id` indicates an audio frames (considering sequence length elements as frames) - `-100` will be ignored in the loss computation - `-101` indicates the audio frame will be used only for the backbone model (using the first codebook token as labels) Such labels can be prepared using `output_labels=True` when calling [`CsmProcessor`]. logits_to_keep (`int` or `torch.Tensor`, *optional*): Kept for compatibility. Does not support another value than: 1. `0`, which is equivalent to keeping all logits, used in the training regime 2. `1`, which is equivalent to keeping only the last logit, used in the generation regime Example: ```python >>> import torch >>> from transformers import CsmForConditionalGeneration, AutoProcessor >>> from datasets import load_dataset, Audio >>> model_id = "sesame/csm-1b" >>> torch_device = "cuda" if torch.cuda.is_available() else "cpu" >>> processor = AutoProcessor.from_pretrained(model_id) >>> ds = load_dataset("hf-internal-testing/dailytalk-dummy", split="train") >>> # ensure the audio is 24kHz >>> ds = ds.cast_column("audio", Audio(sampling_rate=24000)) >>> conversation = [] >>> # prepare a conversation with text and corresponding audio >>> for text, audio, speaker_id in zip(ds[:4]["text"], ds[:4]["audio"], ds[:4]["speaker_id"]): ... conversation.append( ... { ... "role": f"{speaker_id}", ... "content": [{"type": "text", "text": text}, {"type": "audio", "path": audio["array"]}], ... } ... ) >>> inputs = processor.apply_chat_template( ... conversation, ... tokenize=True, ... return_dict=True, ... output_labels=True, ... ).to(torch_device) >>> model = CsmForConditionalGeneration.from_pretrained(model_id, device_map=torch_device) >>> output = model(**inputs) >>> output.loss.backward() ```NrrrI) rrrr%rrrrrr)r$rIrrrrir.r)rTr)rrrrr return_dictrI) r#r-r(r$r%r&r'r)r*r+r,r6)r>rrrrrbr@rrLr^rMrallrKrArrrrdr#r"r%r&r'r$)rOrrrrrr%rrIrrrrrJrrbackbone_outputsbackbone_hidden_statesrNbackbone_logitsr#r-r(depth_decoder_outputsbackbone_labels train_maskdepth_decoder_input_ids train_idxsbackbone_last_hidden_statesdepth_decoder_labelsr6r6r7rrsV   (  z#CsmForConditionalGeneration.forwardrS) NNNNNNNNNNNNr)!r.r/r0rTrarrrjrlro classmethodryrrr2rrrrr3rOrrrrWrrrr=r5r"rrrvr6r6rgr7r]^s     U       r])r8r\rr>r]r4)r;)J dataclassesrtypingrrrr2torch.nnrA activationsr cache_utilsrr generationr integrationsr masking_utilsr modeling_flash_attention_utilsr modeling_layersrmodeling_outputsrrmodeling_rope_utilsrrmodeling_utilsrrprocessing_utilsrutilsrrrrrautorconfiguration_csmrrgeneration_csmr get_loggerr.r$r"r8ModulerMrwrrrrrrrrrr:rrJr=r>rXr\r]__all__r6r6r6r7s            ,!"  F5  h