tjbpdZddlZddlmZddlZddlZddlmZddlm Z ddl m Z ddl mZdd lmZmZmZdd lmZdd lmZmZdd lmZdd lmZddlmZmZmZm Z m!Z!ddl"m#Z#m$Z$ddl%m&Z&ddl'm(Z(m)Z)m*Z*m+Z+m,Z,ddl-m.Z.ddl/m0Z0m1Z1ddl2m3Z3e,j4e5Z6dej7de8de8fdZ9Gddej:Z; d=dej<dej7dej7d ej7d!ej7dzd"e=dzd#e=d$e&e(fd%Z>Gd&d'ej<Z?Gd(d)eZ@Gd*d+eZAe)Gd,d-e$ZBGd.d/eBZCGd0d1eBZDe)Gd2d3eBZEe)d45Gd6d7eBeZFGd8d9eBZGGd:d;eBeZHgd<ZIdS)>z=PyTorch MarianMTModel model, ported from the Marian C++ repo.N)Callable)nn)CrossEntropyLoss)initialization)ACT2FN)Cache DynamicCacheEncoderDecoderCache)GenerationMixin)create_bidirectional_maskcreate_causal_mask)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutput)BaseModelOutputWithPastAndCrossAttentions!CausalLMOutputWithCrossAttentionsSeq2SeqLMOutputSeq2SeqModelOutput)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tupleis_torchdynamo_compilinglogging)merge_with_config_defaults)OutputRecordercapture_outputs) MarianConfig input_ids pad_token_iddecoder_start_token_idc||j}|ddddf|ddddf<||dddf<|td||dk||S)z1 Shift input ids one token to the right. Nr!rz1self.model.config.pad_token_id has to be defined.i) new_zerosshapeclone ValueError masked_fill_)r#r$r%shifted_input_idss h/var/www/html/CA-Chatbot/venv/lib/python3.11/site-packages/transformers/models/marian/modeling_marian.pyshift_tokens_rightr/8s"++IO<<(CRC06688aaae4aaadLMMM""#4#zPMarianSinusoidalPositionalEmbedding.create_weight...Us7 L L LAcBHUAaL3$6777 L L Lr0)range)rCrFrEs @r.rGzEMarianSinusoidalPositionalEmbedding.create_weight..Us9 e e eQT L L L L Ls L L L e e er0F)dtype requires_gradr@rr!N) weightr)rAarrayrHtorchemptyrI FloatTensorsincos)r;n_pos position_encoutsentinelrEs @r. create_weightz1MarianSinusoidalPositionalEmbedding.create_weightNs [& sx e e e eX]^cXdXd e e e  k%DK,=USSS"Qw!||3!88#(a".rvl111add76K/L/LMMAAAqzM!-bf\!!!QTT'5J.K.KLLAAAxyyL r0rinput_ids_shapepast_key_values_length position_idsc|<|dd\}}tj|||ztj|jj}t |S)z3`input_ids_shape` is expected to be [bsz x seqlen].Nr@)rIdevice)rMarangelongrKr[r9forward)r;rWrXrYbszseq_lenr<s r.r^z+MarianSinusoidalPositionalEmbedding.forward]se  *2A2.LC <&(>(HPUPZcgcncuLww|,,,r0N)rN) __name__ __module__ __qualname____doc__intr:rVrMno_gradSizeTensorr^ __classcell__r<s@r.r2r2HsNNEEcE#ECRVJEbfEEEEEE   U]__pt - -$z -CF -Z_ZfimZm -  - - - - -_ - - - - -r0r2modulequerykeyvalueattention_maskscalingdropoutkwargsc ||ddz}tj||dd|z}|||z}tj|d}tj|||j}tj||} | dd } | |fS)Nr'r@rrEptrainingr!) sizerMmatmul transposer functionalsoftmaxrsrz contiguous) rmrnrorprqrrrsrt attn_weights attn_outputs r.eager_attention_forwardrks**R..D(<s}}Q':':;;gEL!#n4 =((2(>>L=((6?([[L,|U33K''1--88::K  $$r0ceZdZdZ ddededed ed ed ed edzd edzffd Z dde j de j dzde dzde j dzde e dee j e j dzff dZxZS)MarianAttentionz=Multi-headed attention from 'Attention Is All You Need' paperrlFTN embed_dim num_headsrs is_decoderbias is_causalconfig layer_idxc zt||_||_||_||z|_||_|j|z|jkrtd|jd|d|jdz|_||_ ||_ ||_ |/|j r(t d|jjdt!j||||_t!j||||_t!j||||_t!j||||_dS)Nz;embed_dim must be divisible by num_heads (got `embed_dim`: z and `num_heads`: z).rvzInstantiating a decoder z without passing `layer_idx` is not recommended and will lead to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` when creating this class.r)r9r:rrrshead_dimrr+rrrrrlogger warning_oncer<rbrLineark_projv_projq_projout_proj) r;rrrsrrrrrr<s r.r:zMarianAttention.__init__sY "" !Y.  MI %$. 8 83dn33%.333 }d* $""      ,4>+B,,,    i 94@@@ i 94@@@ i 94@@@  )YTBBB r0 hidden_stateskey_value_statespast_key_valuesrqrtr6c |du}|jdd}g|d|jR}|||dd} d} |Ht |t r1|j|j } |r|j } n |j } n|} |r|n|} |r3|1| r/| j |j j } | j |j j}n|| } || }g| jddd|jR}| |dd} ||dd}|E| | ||j \} }|r$t |t rd|j|j <t%j|jjt,}||| | ||f|jsdn|j|jd|\}}|jg|dR}||}||fS) z#Input shape: Batch x Time x ChannelNr'r!r@FTrl)rsrr)r)rrviewr} isinstancer is_updatedgetrcross_attention_cacheself_attention_cachelayerskeysvaluesrrupdater get_interfacer_attn_implementationrrzrsrrreshaperr)r;rrrrqrtis_cross_attention input_shape hidden_shape query_statesrcurr_past_key_valuescurrent_states key_states value_stateskv_shapeattention_interfacerrs r.r^zMarianAttention.forwards.T9$)#2#. 88b8$-88 {{=1166|DDNNqRSTT   &/+>?? 7,7;;DNKK %P+:+P((+:+O(('6$-?R))]  F/"=*"=-4T^DIJ/6t~FMLL^44J;;~66LF-crc2FBF FFH#22<(?(M K ,.E) ) %8$7      % $}>CC$,L % %  % % ! \*k);;;;;;FFHH mmK00 L((r0)rlFTFNNNNN)rbrcrdrerffloatboolr"r:rMrir rrtupler^rjrks@r.rrs]GG  &* $%C%C%C%C %C  %C  %C%Ct#%C:%C%C%C%C%C%CT15(,.2 H)H)|H) ,-H) H)  t+ H)-.H) u|U\D00 1H)H)H)H)H)H)H)H)r0rcreZdZd dededzffd Zdejdejdee dej fd Z xZ S) MarianEncoderLayerNrrct|j|_t |j|j|j|||_tj |j|_ |j |_ t|j |_|j|_tj|j|j|_tj|j|j|_tj |j|_dS)N)rrrsrr)r9r:d_modelrrencoder_attention_headsattention_dropout self_attnr LayerNormself_attn_layer_normrsractivation_function activation_fnactivation_dropoutrencoder_ffn_dimfc1fc2final_layer_normr;rrr<s r.r:zMarianEncoderLayer.__init__s (n4,    %'L$@$@!~ #F$>?"(";9T^V-CDD9V3T^DD " T^ < <r0rrqrtr6c |}|j|fd|i|\}}tj||j|j}||z}||}|}|||}tj||j|j}| |}tj||j|j}||z}| |}|j tj kr_tj|s9tj|j jdz }tj|| |}|S)Nrqrxi)minmax)rrr~rsrzrrrrrrrIrMfloat16isfiniteallfinforclamp)r;rrqrtresidual_ clamp_values r.r^zMarianEncoderLayer.forwardsv !)4>   )    q  --mt|VZVc-dd  =0 11-@@  **488M+B+BCC  --mt?Vaean-oo //  --mt|VZVc-dd  =0 --m<<  %- / /}8U8U8Y8Y8[8[ /+m&9::>EK!K K<[YYYMr0ra) rbrcrdr"rfr:rMrOrrrir^rjrks@r.rrs==|=d ======&()+,   r0rceZdZddededzffd Z ddejdejdzdejdzd ejdzd edzd e dzd e e d ejfdZ xZ S)MarianDecoderLayerNrrc t|j|_t |j|j|jdd|||_|j|_t|j |_ |j |_ tj|j|_t |j|j|jd|||_tj|j|_tj|j|j|_tj|j|j|_tj|j|_dS)NT)rrrsrrrr)rsrrr)r9r:rrrdecoder_attention_headsrrrsrrrrrrr encoder_attnencoder_attn_layer_normrdecoder_ffn_dimrrrrs r.r:zMarianDecoderLayer.__init__3s  (n4,   ~ #F$>?"(";$&L$@$@!+ N  *,    (*|DN'C'C$9T^V-CDD9V3T^DD " T^ < <r0Trrqencoder_hidden_statesencoder_attention_maskr use_cachertr6c |}|j|f||d|\}} tj||j|j}||z}||}|]|}|j|f|||d|\}} tj||j|j}||z}||}|}|| |}tj||j |j}| |}tj||j|j}||z}| |}|S)N)rrqrx)rrqr) rrr~rsrzrrrrrrrr) r;rrqrrrrrtrrs r.r^zMarianDecoderLayer.forwardRs!*4>  +)      q  --mt|VZVc-dd  =0 11-@@  ! ,$H0t0 !65 /        M1M11-4.s"$f$f$fA%7%?%?$f$f$fr0F)r9r:rsencoder_layerdrop layerdroprr$r5max_position_embeddingsmax_source_positionsscale_embeddingmathsqrt embed_scaler Embedding vocab_size embed_tokensr2embed_positions ModuleListrHencoder_layersrgradient_checkpointing post_init)r;rrr<s ` r.r:zMarianEncoder.__init__s    ~ 1N !.$*$B!393IR49Y///sL):ItGWXXB  *It7G  m$f$f$f$fvOdIeIe$f$f$fgg &+# r0Nr#rq inputs_embedsrtr6c |du|duz rtd||||jz}||jdd}||z}t j||j|j}t|j ||}t|j D];\}}d} |jr!tjg} | |jkrd} | s |||fi|}.s($s$s$sQR%7!%L%L%L$s$s$sr0F)r9r:rsdecoder_layerdroprr$r5rmax_target_positionsrrrrr rr decoder_vocab_sizer r2r rrHdecoder_layersrrrr;rr<s `r.r:zMarianDecoder.__init__s    ~ 1!.$*$B!8>8NW49V^444TWL)BFNTXTdeeB  *FND?? !O 0 0   );')+    ";;'1"7 " " " ++  $&<<,  & 4  --mt|VZVc-dd "+DK"8"8   C} &+jnn#&77)M%(> /# MM9++    r0)NNNNNNN)rbrcrdrerrrrr"r:rr rrMrrirOr rrrrr^rjrks@r.rrsy,$n_A+VVV*N?!P^___ |& .2.2:>:>(,26!%R R #d*R  t+R  %047 R !& 04 7 R  R (4/R $;R +,R  3R R R ^_ R R R R R r0rc`eZdZddgZdeffd ZdZdZdZdZ d e d e j fd Z ee dd ejd zdejd zdejd zdejd zdeejezd zded zdejd zdejd zded zdeed efdZxZS) MarianModel$model.encoder.embed_positions.weight$model.decoder.embed_positions.weightrcZt||j|j}}|jjr+t j||j||_ ddd|_ nd|_ t||_ t||_|dS)Nz shared.weight)zdecoder.embed_tokens.weightzencoder.embed_tokens.weight)r9r:r$r r share_encoder_decoder_embeddingsrr rshared_tied_weights_keysrencoderrdecoderr)r;rr5r r<s r.r:zMarianModel.__init__ws    "("5v7HZ  ; 7 +,z6>;OODK/>/>''D # # '+D #$V,, $V,,  r0cN|Sra) get_encoderget_input_embeddingsr;s r.rAz MarianModel.get_input_embeddingss !!66888r0c|jjr+||_|j|j_|j|j_dS||j_dSra)rr:r;r=r r>r;rps r.set_input_embeddingsz MarianModel.set_input_embeddingssD ; 7 .DK(, DL %(, DL % % %(-DL % % %r0c|jjrtd|S)Nz`get_decoder_input_embeddings` should not be called if `config.share_encoder_decoder_embeddings` is `True`. Please use `get_input_embeddings` instead.)rr:r+ get_decoderrArBs r.get_decoder_input_embeddingsz(MarianModel.get_decoder_input_embeddingssG ; 7 H !!66888r0cT|jjrtd||j_dS)Na `config.share_encoder_decoder_embeddings` is set to `True` meaning the decoder input embeddings are shared with the encoder. In order to set the decoder input embeddings, you should simply set the encoder input embeddings by calling `set_input_embeddings` with the appropriate embeddings.)rr:r+r>r rDs r.set_decoder_input_embeddingsz(MarianModel.set_decoder_input_embeddingss9 ; 7 r  %* !!!r0new_num_tokensr6c*|jjrtd|}|||}|||}||S||j_||SNz`resize_decoder_token_embeddings` should not be called if `config.share_encoder_decoder_embeddings` is `True`. Please use `resize_token_embeddings` instead.)rr:r+rH_get_resized_embeddingsrJr( tie_weights)r;rKold_embeddingsnew_embeddings model_embedss r.resize_decoder_token_embeddingsz+MarianModel.resize_decoder_token_embeddingss ; 7 K  ::<<55nnUU )).99988::  ! *8 & r0Nr#rqrdecoder_attention_maskencoder_outputsrrdecoder_inputs_embedsrrtc ||jd |||d| }nct|tsNt|dt|dkr|dndt|dkr|dnd}|jd |||d|||| d| } t | j| j| j| j | j |j|j|j S) a decoder_input_ids (`torch.LongTensor` of shape `(batch_size, target_sequence_length)`, *optional*): Indices of decoder input sequence tokens in the vocabulary. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are decoder input IDs?](../glossary#decoder-input-ids) Marian uses the `pad_token_id` as the starting token for `decoder_input_ids` generation. If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`). decoder_attention_mask (`torch.LongTensor` of shape `(batch_size, target_sequence_length)`, *optional*): Default behavior: generate a tensor that ignores pad tokens in `decoder_input_ids`. Causal mask will also be used by default. Example: ```python >>> from transformers import AutoTokenizer, MarianModel >>> tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-de") >>> model = MarianModel.from_pretrained("Helsinki-NLP/opus-mt-en-de") >>> inputs = tokenizer("Studies have been shown that owning a dog is good for you", return_tensors="pt") >>> decoder_inputs = tokenizer( ... " Studien haben gezeigt dass es hilfreich ist einen Hund zu besitzen", ... return_tensors="pt", ... add_special_tokens=False, ... ) >>> outputs = model(input_ids=inputs.input_ids, decoder_input_ids=decoder_inputs.input_ids) >>> last_hidden_states = outputs.last_hidden_state >>> list(last_hidden_states.shape) [1, 26, 512] ```N)r#rqrrr!r@)rrrr#rqrrrrr)rrdecoder_hidden_statesdecoder_attentionsr"encoder_last_hidden_staterencoder_attentionsr) r=rrlenr>rrrrrr") r;r#rqrrTrUrrrVrrtdecoder_outputss r.r^zMarianModel.forwards8h  "*dl#-+ OOO_== -"1!"4474H4H14L4Loa00RV14_1E1E1I1I?1--tO'$,  '1"1!"4#1+/      "-?+;"1"?.9,=&5&G"1"?.9    r0) NNNNNNNNN)rbrcrd_keys_to_ignore_on_load_missingr"r:rArErHrJrfrr rSrrrMrrirrr rOrrrrr^rjrks@r.r6r6ps /.'# |*999...999***cbl0.2.2596:HL(,26:>!%V V #d*V  t+V !+d2 V !& t 3 V u|,>E V V (4/V  %047V $;V +,V  V V V ^V V V V V r0r6zX The Marian Model with a language modeling head. Can be used for summarization. ) custom_introceZdZdZgdZddgZddiZdeffd Z d"d e d e d zd e de j ffd Z d#d e de j fdZdZd e dd fdZde j fdZee d$dejd zdejd zdejd zdejd zdeejezd zded zdejd zdejd zdejd zde d zdeedefd Zdejfd!ZxZ S)%rr)rr7r8r7r8lm_head.weight!model.decoder.embed_tokens.weightrct|t||_|jjr dddd|_|jr|jn|j}| dtj d|ftj |j|d|_|dS)Nzmodel.shared.weight)rbrcz!model.encoder.embed_tokens.weightrr!Fr)r9r:r6rrr:r<r r(register_bufferrMzerosrrrlm_headr)r;rtarget_vocab_sizer<s r.r:zMarianMTModel.__init__+s     (( ; 7 "75J5J''D # 281XwF--^d^w 0%+qBS>T2U2UVVVy1BOOO  r0NTrKpad_to_multiple_of mean_resizingr6ct|||}|jjr|||Sra)r9resize_token_embeddingsrr:_resize_final_logits_bias)r;rKrirjrQr<s r.rlz%MarianMTModel.resize_token_embeddings<sI88I[]jkk ; 7 ;  * *> : : :r0c|}||||}|||jjd}|jjr ||j_|jjr_|K|jj s?|}| ||}| ||S)Nr) rArNrErKr)rr:r(get_output_embeddingstie_word_embeddings_get_resized_lm_headset_output_embeddings)r;rKriargsrPrQ old_lm_head new_lm_heads r._resize_token_embeddingsz&MarianMTModel._resize_token_embeddingsEs224455nnVhii !!.111'.4Q7 ; 7 <-;DK * K 8 4**,,8K394466K33KPPK  & &{ 3 3 3((***r0c0|jjrtd|j}|||}|j||K|jjs?|}| ||}| ||j}||S||j_ | | ||SrM)rr:r+rrHrNrJrorprqrrr(rOrm)r;rKrPrQrtrurRs r.rSz-MarianMTModel.resize_decoder_token_embeddings[s ; 7 K  @@BB55nnUU //???  % % ' ' 3DK<[ 34466K33KPPK  & &{ 3 3 3z>>@@  ! *8 &  &&~666r0c|jjd}||kr|jddd|f}nBtjd||z f|jj}tj|j|gd}|d|dS)Nr'r!rrwr)rr)rMrfr[catre)r;rKold_num_tokensnew_bias extra_biass r.rmz'MarianMTModel._resize_final_logits_bias{s/5b9 ^ + +-aaa..@AHHa.)H%IRVRhRopppJy$"8*!E1MMMH 0(;;;;;r0rQc||_dSra)rg)r;rQs r.rrz#MarianMTModel.set_output_embeddingss % r0r#rqrrTrUrrrVlabelsrrtc | G| rtdd} |'|%t| |jj|jj}|j|f|||||||| d| } || d|jz} d}| Kt}|| d|jj | d}t|| | j | j| j| j| j| j| j S)u decoder_input_ids (`torch.LongTensor` of shape `(batch_size, target_sequence_length)`, *optional*): Indices of decoder input sequence tokens in the vocabulary. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are decoder input IDs?](../glossary#decoder-input-ids) Marian uses the `pad_token_id` as the starting token for `decoder_input_ids` generation. If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`). decoder_attention_mask (`torch.LongTensor` of shape `(batch_size, target_sequence_length)`, *optional*): Default behavior: generate a tensor that ignores pad tokens in `decoder_input_ids`. Causal mask will also be used by default. 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]`. Example: ```python >>> from transformers import AutoTokenizer, MarianMTModel >>> src = "fr" # source language >>> trg = "en" # target language >>> model_name = f"Helsinki-NLP/opus-mt-{src}-{trg}" >>> model = MarianMTModel.from_pretrained(model_name) >>> tokenizer = AutoTokenizer.from_pretrained(model_name) >>> sample_text = "où est l'arrêt de bus ?" >>> batch = tokenizer([sample_text], return_tensors="pt") >>> generated_ids = model.generate(**batch) >>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0] "Where's the bus stop?" ``` NzJThe `use_cache` argument is changed to `False` since `labels` is provided.F)rqrrUrTrrrVrrr') losslogitsrrYrZr"r[rr\)rwarningr/rr$r%rrgrrrr(rrrYrZr"r[rr\)r;r#rqrrTrUrrrVr~rrtoutputs lm_logitsmasked_lm_lossloss_fcts r.r^zMarianMTModel.forwardsMp   mklllI (-B-J$6DK4dk6X%%!'1dj  ' )/+#9+'"7 ' '  ' ' LL,,t/EE   '))H%XinnR9W&X&XZ`ZeZefhZiZijjN#3")"?&9$5&-&G")"?&9    r0cLt||jj|jjSra)r/rr$r%)r;r~s r.%prepare_decoder_input_ids_from_labelsz3MarianMTModel.prepare_decoder_input_ids_from_labelss!&$+*BDKDfgggr0)NTra) NNNNNNNNNN)!rbrcrdrr__keys_to_ignore_on_saver<r"r:rfrrr rlrvrSrmrrrrrMrrirrr rOrrrr^rrjrks@r.rrs  '''# FGmn*,OP|$ae!7:TzY] ++s+_a_k++++,@<<<<<<&BL&&&&.2.2596:HL(,26:>*.!%] ] #d*]  t+] !+d2 ] !& t 3 ] u|,>E ] ] (4/]  %047]  4'] $;] +,]  ] ] ] ^] ~hELhhhhhhhhr0rc(eZdZdZfdZdZxZS)MarianDecoderWrapperz This wrapper class is a helper class to correctly load pretrained checkpoints when the causal language model is used in combination with the [`EncoderDecoderModel`] framework. ct|t||_|dSra)r9r:rr>rr*s r.r:zMarianDecoderWrapper.__init__s@    $V,,  r0c|j|i|Sra)r>)r;rsrts r.r^zMarianDecoderWrapper.forwardst|T,V,,,r0)rbrcrdrer:r^rjrks@r.rrsQ  -------r0rc(eZdZddiZfdZdZdZee dde j dzd e j dzd e j dzd e j dzd e dzd e j dzde j dzdedzdee j zdeedeezfdZxZS)MarianForCausalLMrbrccd|_d|_t|t ||_t j|j|j d|_ | dS)NTFr) rr,r9r:rrrr hidden_sizer rgrr*s r.r:zMarianForCausalLM.__init__sp $)!    )&11 y!3V5FUSSS  r0c$|jjjSrarr>r rBs r.rAz&MarianForCausalLM.get_input_embeddings sz!..r0c(||jj_dSrarrDs r.rEz&MarianForCausalLM.set_input_embeddingss*/ '''r0Nrr#rqrrrrr~rlogits_to_keeprtr6c |jjd|||||||d| } | d} t| trt | dn| } || dd| ddf}d}|e||j}t}|| d|j j | d}t||| j | j| j| jS)aT 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]`. Example: ```python >>> from transformers import AutoTokenizer, MarianForCausalLM >>> tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-fr-en") >>> model = MarianForCausalLM.from_pretrained("Helsinki-NLP/opus-mt-fr-en") >>> assert model.config.is_decoder, f"{model.__class__} has to be configured as a decoder." >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt") >>> outputs = model(**inputs) >>> logits = outputs.logits >>> expected_shape = [1, inputs.input_ids.shape[-1], model.config.vocab_size] >>> list(logits.shape) == expected_shape True ```rXrNr')rrrrrr"r)rr>rrfslicergtor[rrrr rrrrr")r;r#rqrrrrr~rrrtrr slice_indicesrrrs r.r^zMarianForCausalLM.forwards)L>PTZ=O > )"7#9+' > >  > >   8B>SV8W8Wk~ot444]k mAAA}aaa,?@AA  YYv}--F'))H8FKKDK,BCCV[[QS__UUD0#3!/)$5     r0) NNNNNNNNr)rbrcrdr<r:rArErrrMrrirOr rrfrrrrr^rjrks@r.rrs{=     ///000.2.2:>;?(,26*.!%-.A A #d*A  t+A  %047 A !& 1D 8 A  A (4/A  4'A $;A el*A +,A  2 2A A A ^A A A A A r0r)rr6rr)Nrl)Jrercollections.abcrnumpyrArMrtorch.nnrrr activationsr cache_utilsr r r generationr masking_utilsr rmodeling_flash_attention_utilsrmodeling_layersrmodeling_outputsrrrrrmodeling_utilsrrprocessing_utilsrutilsrrrrr utils.genericrutils.output_capturingrr configuration_marianr" get_loggerrbrrirfr/r r2Modulerrrrrrrrr6rrr__all__rr0r.rs%DC $$$$$$ %%%%%%&&&&&&!!!!!!CCCCCCCCCC))))))JJJJJJJJBBBBBB999999GFFFFFFF&&&&&&877777EEEEEEEE......  H % % %, c [^     -----",---R!%% I% <% % < % L4' % T\ %%'(%%%%:r)r)r)r)r)bir)r)r)l000003000hOOOOO3OOOdO