Ë
    ÜÓ^jqW  ã                   ó˜  — d Z ddlmZ ddlmZ ddlZddlmc mZ	 ddlm
Z
mZmZ ddlmZ ddlmZ dd	lmZmZ dd
lmZ ddlmZ ddlmZmZmZmZ ddlmZmZ ddl m!Z!m"Z" ddl#m$Z$m%Z% ddl&m'Z' ddl(m)Z)m*Z*m+Z+m,Z, ddl-m.Z.m/Z/ ddl0m1Z1 ddl2m3Z3  e,jh                  e5«      Z6d„ Z7 G d„ dejp                  «      Z9 G d„ dejt                  «      Z;d„ Z<d=d„Z= G d„ dejt                  «      Z>dej                  d e?d!ej                  fd"„Z@	 d>d#ejt                  d$ej                  d%ej                  d&ej                  d'ej                  dz  d(eAd)eAd*e'e)   fd+„ZB G d,„ d-ejt                  «      ZC G d.„ d/e«      ZDe* G d0„ d1e%«      «       ZEe* G d2„ d3eE«      «       ZF G d4„ d5eEe«      ZG G d6„ d7eeE«      ZH G d8„ d9eeE«      ZI G d:„ d;eeE«      ZJg d<¢ZKy)?zPyTorch Nemotron model.é    )ÚCallable)ÚOptionalN)ÚSizeÚTensorÚnné   )Úinitialization)ÚACT2FN)ÚCacheÚDynamicCache)ÚGenerationMixin)Úcreate_causal_mask)ÚGenericForQuestionAnsweringÚ GenericForSequenceClassificationÚGenericForTokenClassificationÚGradientCheckpointingLayer)ÚBaseModelOutputWithPastÚCausalLMOutputWithPast)ÚROPE_INIT_FUNCTIONSÚdynamic_rope_update)ÚALL_ATTENTION_FUNCTIONSÚPreTrainedModel)ÚUnpack)ÚTransformersKwargsÚauto_docstringÚcan_return_tupleÚlogging)Úmaybe_autocastÚmerge_with_config_defaults)Úcapture_outputsé   )ÚNemotronConfigc                 ó®   — t        j                  «       s|S t        j                  | «      }t         j                  j                  j                  || |«      S ©N)ÚtorchÚis_autocast_enabledÚget_autocast_dtypeÚampÚautocast_modeÚ_cast)Údevice_typeÚargsÚtarget_dtypes      úy/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/nemotron/modeling_nemotron.pyÚ_cast_if_autocast_enabledr/   6   sB   € Ü×$Ñ$Ô&Øˆä×/Ñ/°Ó<ˆÜ�y‰y×&Ñ&×,Ñ,¨T°;ÀÓMÐMó    c            	       ó^   ‡ — e Zd Z	 	 	 	 	 d	deee   z  ez  dededefˆ fd„Zde	de	fd„Z
ˆ xZS )
ÚNemotronLayerNorm1PÚnormalized_shapeÚepsÚelementwise_affineÚbiasc                 ó.   •— t         ‰| �  ||||||«       y r$   )ÚsuperÚ__init__)Úselfr3   r4   r5   r6   ÚdeviceÚdtypeÚ	__class__s          €r.   r9   zNemotronLayerNorm1P.__init__?   s   ø€ ô 	‰ÑÐ)¨3Ð0BÀDÈ&ÐRWÕXr0   ÚinputÚreturnc                 ól  — |j                   j                  dk7  r|j                   j                  nd}t        ||| j                  | j                  dz   | j
                  | j                  «      }t        |j                   j                  d¬«      5  t        j                  |Ž cd d d «       S # 1 sw Y   y xY w)NÚmpsÚcpuç      ð?F©r+   Úenabled)
r;   Útyper/   r3   Úweightr6   r4   r   ÚFÚ
layer_norm)r:   r>   r+   r,   s       r.   ÚforwardzNemotronLayerNorm1P.forwardJ   sŒ   € Ø+0¯<©<×+<Ñ+<ÀÒ+E�e—l‘l×'Ò'È5ˆÜ(Ø˜ × 5Ñ 5°t·{±{ÀSÑ7HÈ$Ï)É)ÐUY×U]ÑU]ó
ˆô ¨¯©×(9Ñ(9À5ÔIñ 	'Ü—<‘< Ð&÷	'÷ 	'ò 	'ús   ÂB*Â*B3)gñhãˆµøä>TTNN)Ú__name__Ú
__module__Ú__qualname__ÚintÚlistr   ÚfloatÚboolr9   r   rJ   Ú__classcell__©r=   s   @r.   r2   r2   >   sd   ø„ ð Ø#'ØØØñ	Yà  S¡	™/¨DÑ0ð	Yð ð	Yð !ð		Yð
 õ	Yð'˜Vð '¨÷ 'r0   r2   c                   óÄ   ‡ — e Zd ZU ej                  ed<   ddefˆ fd„Ze	 	 	 ddedz  de	d   de
dz  ded	ef   fd
„«       Z ej                  «       ed„ «       «       Zˆ xZS )ÚNemotronRotaryEmbeddingÚinv_freqNÚconfigc                 ó®  •— t         ‰| �  «        |j                  | _        |j                  | _        || _        | j
                  j                  d   | _        | j                  }| j                  dk7  rt        | j                     } || j
                  |«      \  }| _
        | j                  d|d¬«       | j                  d|j                  «       d¬«       y )NÚ	rope_typeÚdefaultrV   F)Ú
persistentÚoriginal_inv_freq)r8   r9   Úmax_position_embeddingsÚmax_seq_len_cachedÚoriginal_max_seq_lenrW   Úrope_parametersrY   Úcompute_default_rope_parametersr   Úattention_scalingÚregister_bufferÚclone)r:   rW   r;   Úrope_init_fnrV   r=   s        €r.   r9   z NemotronRotaryEmbedding.__init__W   s¶   ø€ Ü‰ÑÔØ"(×"@Ñ"@ˆÔØ$*×$BÑ$BˆÔ!àˆŒàŸ™×4Ñ4°[ÑAˆŒØ!%×!EÑ!EˆØ�>‰>˜YÒ&Ü.¨t¯~©~Ñ>ˆLÙ+7¸¿¹ÀVÓ+LÑ(ˆ�$Ô(à×Ñ˜Z¨¸eÐÔDØ×ÑÐ0°(·.±.Ó2BÈuÐÕUr0   r;   ztorch.deviceÚseq_lenr?   ztorch.Tensorc                 ón  — | j                   d   }| j                   j                  dd«      }t        | dd«      xs | j                  | j                  z  }t        ||z  «      }d}d|t        j                  d|dt        j                  ¬«      j                  |t        j                  ¬	«      |z  z  z  }||fS )
a¨  
        Computes the inverse frequencies according to the original RoPE implementation
        Args:
            config ([`~transformers.PreTrainedConfig`]):
                The model configuration.
            device (`torch.device`):
                The device to use for initialization of the inverse frequencies.
            seq_len (`int`, *optional*):
                The current sequence length. Unused for this type of RoPE.
        Returns:
            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
        Ú
rope_thetaÚpartial_rotary_factorrC   Úhead_dimNr   é   ©r<   )r;   r<   )r`   ÚgetÚgetattrÚhidden_sizeÚnum_attention_headsrN   r%   ÚarangeÚint64ÚtorP   )	rW   r;   rf   Úbaseri   rj   ÚdimÚattention_factorrV   s	            r.   ra   z7NemotronRotaryEmbedding.compute_default_rope_parametersg   s»   € ð( ×%Ñ% lÑ3ˆØ &× 6Ñ 6× :Ñ :Ð;RÐTWÓ XÐÜ˜6 :¨tÓ4Òh¸×8JÑ8JÈf×NhÑNhÑ8hˆÜ�(Ð2Ñ2Ó3ˆàÐð Ø”U—\‘\ ! S¨!´5·;±;Ô?×BÑBÈ&ÔX]×XcÑXcÐBÓdÐgjÑjÑkñ
ˆð Ð)Ð)Ð)r0   c                 óN  — | j                   d d d …d f   j                  «       j                  |j                  d   dd«      j	                  |j
                  «      }|d d …d d d …f   j                  «       }t        |j
                  j                  t        «      r/|j
                  j                  dk7  r|j
                  j                  nd}t        |d¬«      5  |j                  «       |j                  «       z  j                  dd«      }t        j                  ||fd¬	«      }|j                  «       | j                  z  }|j                  «       | j                  z  }	d d d «       j	                  |j                   ¬
«      	j	                  |j                   ¬
«      fS # 1 sw Y   ŒAxY w)Nr   éÿÿÿÿr!   rA   rB   FrD   rk   ©ru   rl   )rV   rP   ÚexpandÚshapers   r;   Ú
isinstancerF   Ústrr   Ú	transposer%   ÚcatÚcosrb   Úsinr<   )
r:   ÚxÚposition_idsÚinv_freq_expandedÚposition_ids_expandedr+   ÚfreqsÚembr€   r�   s
             r.   rJ   zNemotronRotaryEmbedding.forwardˆ   sR  € ð !ŸM™M¨$²°4¨-Ñ8×>Ñ>Ó@×GÑGÈ×HZÑHZÐ[\ÑH]Ð_aÐcdÓe×hÑhÐij×iqÑiqÓrÐØ ,ªQ°²a¨ZÑ 8× >Ñ >Ó @Ðä'1°!·(±(·-±-ÄÔ'EÈ!Ï(É(Ï-É-Ð[`ÒJ`�a—h‘h—m’mÐfkˆÜ¨¸UÔCñ 	5Ø&×,Ñ,Ó.Ð1F×1LÑ1LÓ1NÑN×YÑYÐZ[Ð]^Ó_ˆEÜ—)‘)˜U E˜N°Ô3ˆCØ—'‘'“)˜d×4Ñ4Ñ4ˆCØ—'‘'“)˜d×4Ñ4Ñ4ˆC÷		5ð �v‰v˜AŸG™GˆvÓ$ c§f¡f°1·7±7 fÓ&;Ð;Ð;÷	5ð 	5ús   ÃBFÆF$r$   )NNN)rK   rL   rM   r%   r   Ú__annotations__r"   r9   Ústaticmethodr   rN   ÚtuplerP   ra   Úno_gradr   rJ   rR   rS   s   @r.   rU   rU   T   sš   ø… Ø�l‰lÓñV˜~õ Vð  ð )-Ø+/Ø"ñ*Ø Ñ%ð*à˜Ñ(ð*ð �t‘ð*ð 
ˆ~˜uÐ$Ñ	%ò	*ó ð*ð> €U‡]�]ƒ_Øñ<ó ó ô<r0   rU   c                 óš   — | dd| j                   d   dz  …f   }| d| j                   d   dz  d…f   }t        j                  | |fd¬«      S )z*Rotates half the hidden dims of the input..Nrx   rk   ry   )r{   r%   r   )r‚   Úx1Úx2s      r.   Úrotate_halfr�   ™   sZ   € à	
ˆ3Ð"�!—'‘'˜"‘+ Ñ"Ð"Ð"Ñ	#€BØ	
ˆ3�—‘˜‘˜qÑ Ñ"Ð"Ñ	#€BÜ�9‰9�r�c˜2�Y BÔ'Ð'r0   c                 ó`  — |j                  |«      }|j                  |«      }|j                  d   }| dd|…f   | d|d…f   }} |dd|…f   |d|d…f   }}| |z  t        | «      |z  z   }||z  t        |«      |z  z   }	t        j                  ||fd¬«      t        j                  |	|fd¬«      fS )a…  Applies 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.
        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.
    rx   .Nry   )Ú	unsqueezer{   r�   r%   r   )
ÚqÚkr€   r�   Úunsqueeze_dimÚrot_dimÚq_passÚk_passÚq_embedÚk_embeds
             r.   Úapply_rotary_pos_embrš       sÉ   € ð$ �-‰-˜Ó
&€CØ
�-‰-˜Ó
&€Cà�i‰i˜‰m€Gà�#�x˜�x�-Ñ  ! C¨© MÑ"2€v€AØ�#�x˜�x�-Ñ  ! C¨© MÑ"2€v€Aà�3‰wœ; q›>¨CÑ/Ñ0€GØ�3‰wœ; q›>¨CÑ/Ñ0€GÜ�9‰9�g˜vÐ&¨BÔ/´·±¸GÀVÐ;LÐRTÔ1UÐUÐUr0   c                   ó$   ‡ — e Zd Zˆ fd„Zd„ Zˆ xZS )ÚNemotronMLPc                 ó�  •— t         ‰| �  «        || _        |j                  | _        |j                  | _        t        j                  | j                  | j                  |j                  ¬«      | _        t        j                  | j                  | j                  |j                  ¬«      | _	        t        |j                     | _        y )N©r6   )r8   r9   rW   ro   Úintermediate_sizer   ÚLinearÚmlp_biasÚup_projÚ	down_projr
   Ú
hidden_actÚact_fn©r:   rW   r=   s     €r.   r9   zNemotronMLP.__init__À   s‘   ø€ Ü‰ÑÔØˆŒØ!×-Ñ-ˆÔØ!'×!9Ñ!9ˆÔÜ—y‘y ×!1Ñ!1°4×3IÑ3IÐPV×P_ÑP_Ô`ˆŒÜŸ™ 4×#9Ñ#9¸4×;KÑ;KÐRX×RaÑRaÔbˆŒÜ˜V×.Ñ.Ñ/ˆ�r0   c                 ó`   — | j                  | j                  | j                  |«      «      «      S r$   )r£   r¥   r¢   )r:   r‚   s     r.   rJ   zNemotronMLP.forwardÉ   s"   € Ø�~‰~˜dŸk™k¨$¯,©,°q«/Ó:Ó;Ð;r0   )rK   rL   rM   r9   rJ   rR   rS   s   @r.   rœ   rœ   ¿   s   ø„ ô0ö<r0   rœ   Úhidden_statesÚn_repr?   c                 óª   — | j                   \  }}}}|dk(  r| S | dd…dd…ddd…dd…f   j                  |||||«      } | j                  |||z  ||«      S )zÔ
    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)
    r!   N)r{   rz   Úreshape)r¨   r©   ÚbatchÚnum_key_value_headsÚslenrj   s         r.   Ú	repeat_kvr¯   Î   so   € ð
 2?×1DÑ1DÑ.€EÐ  hØ�‚zØÐØ!¢!¢Q¨ªa²Ð"2Ñ3×:Ñ:¸5ÐBUÐW\Ð^bÐdlÓm€MØ× Ñ  Ð(;¸eÑ(CÀTÈ8ÓTÐTr0   ÚmoduleÚqueryÚkeyÚvalueÚattention_maskÚscalingÚdropoutÚkwargsc                 ó  — t        || j                  «      }t        || j                  «      }	t        j                  ||j	                  dd«      «      |z  }
|�|
|z   }
t
        j                  j                  |
dt        j                  ¬«      j                  |j                  «      }
t
        j                  j                  |
|| j                  ¬«      }
t        j                  |
|	«      }|j	                  dd«      j                  «       }||
fS )Nrk   r   rx   )ru   r<   )ÚpÚtrainingr!   )r¯   Únum_key_value_groupsr%   Úmatmulr~   r   Ú
functionalÚsoftmaxÚfloat32rs   r<   r¶   rº   Ú
contiguous)r°   r±   r²   r³   r´   rµ   r¶   r·   Ú
key_statesÚvalue_statesÚattn_weightsÚattn_outputs               r.   Úeager_attention_forwardrÅ   Û   sà   € ô ˜3 × ;Ñ ;Ó<€JÜ˜U F×$?Ñ$?Ó@€Lä—<‘<  z×';Ñ';¸A¸qÓ'AÓBÀWÑL€LØÐ!Ø# nÑ4ˆä—=‘=×(Ñ(¨¸2ÄUÇ]Á]Ð(ÓS×VÑVÐW\×WbÑWbÓc€LÜ—=‘=×(Ñ(¨¸È6Ï?É?Ð(Ó[€LÜ—,‘,˜|¨\Ó:€KØ×'Ñ'¨¨1Ó-×8Ñ8Ó:€Kà˜Ð$Ð$r0   c                   óø   ‡ — e Zd ZdZddededz  fˆ fd„Z	 	 ddej                  de	ej                  ej                  f   dej                  dz  d	e
dz  d
ee   de	ej                  ej                  f   fd„Zˆ xZS )ÚNemotronAttentionz=Multi-headed attention from 'Attention Is All You Need' paperNrW   Ú	layer_idxc                 óÂ  •— t         ‰| �  «        || _        || _        |j                  | _        |j
                  | _        |j                  | _        |j                  | _        |j                  | _	        | j                  | j                  z  | _
        | j                  dz  | _        |j                  d   | _        d| _        t        j                   | j
                  | j                  | j                  z  |j"                  ¬«      | _        t        j                   | j
                  | j                  | j                  z  |j"                  ¬«      | _        t        j                   | j
                  | j                  | j                  z  |j"                  ¬«      | _        t        j                   | j                  | j                  z  | j
                  |j"                  ¬«      | _        y )Ng      à¿ri   Trž   )r8   r9   rW   rÈ   Úattention_dropoutro   rp   Ú	num_headsrj   r­   r»   rµ   r`   ri   Ú	is_causalr   r    Úattention_biasÚq_projÚk_projÚv_projÚo_proj©r:   rW   rÈ   r=   s      €r.   r9   zNemotronAttention.__init__÷   s\  ø€ Ü‰ÑÔØˆŒØ"ˆŒà!'×!9Ñ!9ˆÔØ!×-Ñ-ˆÔØ×3Ñ3ˆŒØŸ™ˆŒØ#)×#=Ñ#=ˆÔ Ø$(§N¡N°d×6NÑ6NÑ$NˆÔ!Ø—}‘} dÑ*ˆŒà%+×%;Ñ%;Ð<SÑ%TˆÔ"ØˆŒä—i‘i × 0Ñ 0°$·.±.À4Ç=Á=Ñ2PÐW]×WlÑWlÔmˆŒÜ—i‘i × 0Ñ 0°$×2JÑ2JÈTÏ]É]Ñ2ZÐag×avÑavÔwˆŒÜ—i‘i × 0Ñ 0°$×2JÑ2JÈTÏ]É]Ñ2ZÐag×avÑavÔwˆŒÜ—i‘i §¡°·±Ñ >À×@PÑ@PÐW]×WlÑWlÔmˆ�r0   r¨   Úposition_embeddingsr´   Úpast_key_valuesr·   r?   c                 ó
  — |j                   d d }g |¢d‘| j                  ‘­}| j                  |«      j                  |«      j	                  dd«      }| j                  |«      j                  |«      j	                  dd«      }	| j                  |«      j                  |«      j	                  dd«      }
|\  }}t        ||	||«      \  }}	|� |j                  |	|
| j                  «      \  }	}
t        j                  | j                  j                  t        «      } || ||	|
|f| j                  sdn| j                   | j"                  dœ|¤Ž\  }} |j$                  g |¢d‘­Ž j'                  «       }| j)                  |«      }||fS )Nrx   r!   rk   ç        )r¶   rµ   )r{   rj   rÎ   Úviewr~   rÏ   rÐ   rš   ÚupdaterÈ   r   Úget_interfacerW   Ú_attn_implementationrÅ   rº   rÊ   rµ   r«   rÀ   rÑ   )r:   r¨   rÓ   r´   rÔ   r·   Úinput_shapeÚhidden_shapeÚquery_statesrÁ   rÂ   r€   r�   Úattention_interfacerÄ   rÃ   s                   r.   rJ   zNemotronAttention.forward  s‘  € ð $×)Ñ)¨#¨2Ð.ˆØ8˜Ð8 bÐ8¨$¯-©-Ñ8ˆà—{‘{ =Ó1×6Ñ6°|ÓD×NÑNÈqÐRSÓTˆØ—[‘[ Ó/×4Ñ4°\ÓB×LÑLÈQÐPQÓRˆ
Ø—{‘{ =Ó1×6Ñ6°|ÓD×NÑNÈqÐRSÓTˆà&‰ˆˆSÜ#7¸ÀjÐRUÐWZÓ#[Ñ ˆ�jàÐ&Ø'6×'=Ñ'=¸jÈ,ÐX\×XfÑXfÓ'gÑ$ˆJ˜ä(?×(MÑ(MØ�K‰K×,Ñ,Ô.Eó)
Ðñ %8ØØØØØð	%
ð  $Ÿ}š}‘C°$×2HÑ2HØ—L‘Lñ	%
ð ñ	%
Ñ!ˆ�\ð *�k×)Ñ)Ð;¨;Ð;¸Ò;×FÑFÓHˆØ—k‘k +Ó.ˆØ˜LÐ(Ð(r0   r$   )NN)rK   rL   rM   Ú__doc__r"   rN   r9   r%   r   rŠ   r   r   r   rJ   rR   rS   s   @r.   rÇ   rÇ   ô   s¢   ø„ ÙGñn˜~ð n¸#À¹*õ nð2 /3Ø(,ñ&)à—|‘|ð&)ð # 5§<¡<°·±Ð#=Ñ>ð&)ð Ÿ™ tÑ+ð	&)ð
  ™ð&)ð Ð+Ñ,ð&)ð 
ˆu�|‰|˜UŸ\™\Ð)Ñ	*÷&)r0   rÇ   c                   óø   ‡ — e Zd Zdedefˆ fd„Z	 	 	 	 	 ddej                  dej                  dz  dej                  dz  de	dz  d	e
dz  d
eej                  ej                  f   dz  dej                  fd„Zˆ xZS )ÚNemotronDecoderLayerrW   rÈ   c                 ó   •— t         ‰| �  «        |j                  | _        t        ||¬«      | _        t        |«      | _        t        |j                  |j                  ¬«      | _	        t        |j                  |j                  ¬«      | _
        y )N)rW   rÈ   ©r4   )r8   r9   ro   rÇ   Ú	self_attnrœ   Úmlpr2   Únorm_epsÚinput_layernormÚpost_attention_layernormrÒ   s      €r.   r9   zNemotronDecoderLayer.__init__6  sk   ø€ Ü‰ÑÔØ!×-Ñ-ˆÔä*°&ÀIÔNˆŒä˜vÓ&ˆŒÜ2°6×3EÑ3EÈ6Ï?É?Ô[ˆÔÜ(;¸F×<NÑ<NÐTZ×TcÑTcÔ(dˆÕ%r0   Nr¨   r´   rƒ   rÔ   Ú	use_cacherÓ   r?   c                 ó¼   — |}| j                  |«      }| j                  ||||||¬«      \  }}	||z   }|}| j                  |«      }| j                  |«      }||z   }|S )N)r¨   r´   rƒ   rÔ   ré   rÓ   )rç   rä   rè   rå   )
r:   r¨   r´   rƒ   rÔ   ré   rÓ   r·   ÚresidualÚ_s
             r.   rJ   zNemotronDecoderLayer.forward@  s†   € ð !ˆà×,Ñ,¨]Ó;ˆð  Ÿ>™>Ø'Ø)Ø%Ø+ØØ 3ð *ó 
Ñˆ�qð ! =Ñ0ˆð !ˆØ×5Ñ5°mÓDˆØŸ™ Ó/ˆØ  =Ñ0ˆàÐr0   )NNNFN)rK   rL   rM   r"   rN   r9   r%   r   Ú
LongTensorr   rQ   rŠ   rJ   rR   rS   s   @r.   rá   rá   5  s®   ø„ ðe˜~ð e¸#õ eð /3Ø04Ø(,Ø!&ØHLñ à—|‘|ð ð Ÿ™ tÑ+ð ð ×&Ñ&¨Ñ-ð	 ð
  ™ð ð ˜$‘;ð ð # 5§<¡<°·±Ð#=Ñ>ÀÑEð ð 
�‰÷ r0   rá   c                   ó„   ‡ — e Zd ZU eed<   dZdZdgZdgZdZ	dZ
dZdZdZeedœZ ej$                  «       ˆ fd„«       Zˆ xZS )ÚNemotronPreTrainedModelrW   ÚmodelTrá   rÔ   )r¨   Ú
attentionsc                 óÂ   •— t         ‰| �  |«       t        |t        «      r?t	        j
                  |j                  «       t	        j                  |j                  «       y y r$   )	r8   Ú_init_weightsr|   r2   ÚinitÚones_rG   Úzeros_r6   )r:   r°   r=   s     €r.   ró   z%NemotronPreTrainedModel._init_weightsu  s@   ø€ ä‰Ñ˜fÔ%Ü�fÔ1Ô2Ü�J‰J�v—}‘}Ô%Ü�K‰K˜Ÿ™Õ$ð 3r0   )rK   rL   rM   r"   rˆ   Úbase_model_prefixÚsupports_gradient_checkpointingÚ_no_split_modulesÚ_skip_keys_device_placementÚ_supports_flash_attnÚ_supports_sdpaÚ_supports_flex_attnÚ_supports_attention_backendÚ_can_compile_fullgraphrá   rÇ   Ú_can_record_outputsr%   r‹   ró   rR   rS   s   @r.   rï   rï   c  sn   ø… àÓØÐØ&*Ð#Ø/Ð0ÐØ#4Ð"5ÐØÐØ€NØÐØ"&Ðà!Ðà-Ø'ñÐð
 €U‡]�]ƒ_ó%ó ô%r0   rï   c                   óö   ‡ — e Zd ZdZdefˆ fd„Zeee	 	 	 	 	 	 dde	j                  dz  de	j                  dz  de	j                  dz  dedz  d	e	j                  dz  d
edz  dee   defd„«       «       «       Zˆ xZS )ÚNemotronModelz¡
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`NemotronDecoderLayer`]

    Args:
        config: NemotronConfig
    rW   c           	      ó  •— t         ‰| �  |«       |j                  | _        |j                  | _        t        j                  |j                  |j                  | j                  «      | _        t        j                  t        |j                  «      D �cg c]  }t        ||«      ‘Œ c}«      | _        t        |j                  |j                  ¬«      | _        t#        |¬«      | _        d| _        | j)                  «        y c c}w )Nrã   ©rW   F)r8   r9   Úpad_token_idÚpadding_idxÚ
vocab_sizer   Ú	Embeddingro   Úembed_tokensÚ
ModuleListÚrangeÚnum_hidden_layersrá   Úlayersr2   ræ   ÚnormrU   Ú
rotary_embÚgradient_checkpointingÚ	post_initrÒ   s      €r.   r9   zNemotronModel.__init__†  sÄ   ø€ Ü‰Ñ˜Ô Ø!×.Ñ.ˆÔØ ×+Ñ+ˆŒäŸL™L¨×):Ñ):¸F×<NÑ<NÐPT×P`ÑP`ÓaˆÔÜ—m‘mÜFKÈF×LdÑLdÓFeÖf¸Ô! &¨)Õ4Òfó
ˆŒô (¨×(:Ñ(:ÀÇÁÔPˆŒ	Ü1¸Ô@ˆŒØ&+ˆÔ#ð 	�‰Õùò gs   ÂDNÚ	input_idsr´   rƒ   rÔ   Úinputs_embedsré   r·   r?   c           
      ó  — |d u |d uz  rt        d«      ‚|r|€t        | j                  ¬«      }|€| j                  |«      }|€V|�|j	                  «       nd}t        j                  |j                  d   |j                  ¬«      |z   }|j                  d«      }t        | j                  ||||¬«      }	|}
| j                  |
|¬«      }| j                  D ]  } ||
f|	||||dœ|¤Ž}
Œ | j                  |
«      }
t        |
|¬	«      S )
Nz:You must specify exactly one of input_ids or inputs_embedsr  r   r!   )r;   )rW   r  r´   rÔ   rƒ   )rƒ   )r´   rƒ   rÔ   ré   rÓ   )Úlast_hidden_staterÔ   )Ú
ValueErrorr   rW   r	  Úget_seq_lengthr%   rq   r{   r;   r‘   r   r  r  r  r   )r:   r  r´   rƒ   rÔ   r  ré   r·   Úpast_seen_tokensÚcausal_maskr¨   rÓ   Údecoder_layers                r.   rJ   zNemotronModel.forward–  s:  € ð ˜Ð -°tÐ";Ò<ÜÐYÓZÐZá˜Ð0Ü*°$·+±+Ô>ˆOàÐ Ø ×-Ñ-¨iÓ8ˆMàÐØCRÐC^˜×=Ñ=Ô?ÐdeÐÜ Ÿ<™<¨×(;Ñ(;¸AÑ(>À}×G[ÑG[Ô\Ð_oÑoˆLØ'×1Ñ1°!Ó4ˆLä(Ø—;‘;Ø'Ø)Ø+Ø%ô
ˆð &ˆØ"Ÿo™o¨mÈ,˜oÓWÐà!Ÿ[™[ò 		ˆMÙ)Øðà*Ø)Ø /Ø#Ø$7ñð ñ‰Mð		ð Ÿ	™	 -Ó0ˆä&Ø+Ø+ô
ð 	
r0   )NNNNNN)rK   rL   rM   rß   r"   r9   r   r    r   r%   rí   r   r   ÚFloatTensorrQ   r   r   r   rJ   rR   rS   s   @r.   r  r  }  sÒ   ø„ ñð˜~õ ð   ØØð .2Ø.2Ø04Ø(,Ø26Ø!%ñ3
à×#Ñ# dÑ*ð3
ð Ÿ™ tÑ+ð3
ð ×&Ñ&¨Ñ-ð	3
ð
  ™ð3
ð ×(Ñ(¨4Ñ/ð3
ð ˜$‘;ð3
ð Ð+Ñ,ð3
ð 
!ò3
ó ó ó  ô3
r0   r  c                   ó*  ‡ — e Zd ZddiZˆ fd„Zee	 	 	 	 	 	 	 	 ddej                  dz  dej                  dz  dej                  dz  de
dz  d	ej                  dz  d
ej                  dz  dedz  deej                  z  dee   defd„«       «       Zˆ xZS )ÚNemotronForCausalLMzlm_head.weightzmodel.embed_tokens.weightc                 óè   •— t         ‰| �  |«       t        |«      | _        |j                  | _        t        j                  |j                  |j                  d¬«      | _        | j                  «        y )NFrž   )
r8   r9   r  rð   r  r   r    ro   Úlm_headr  r¦   s     €r.   r9   zNemotronForCausalLM.__init__Ó  sU   ø€ Ü‰Ñ˜Ô Ü" 6Ó*ˆŒ
Ø ×+Ñ+ˆŒÜ—y‘y ×!3Ñ!3°V×5FÑ5FÈUÔSˆŒð 	�‰Õr0   Nr  r´   rƒ   rÔ   r  Úlabelsré   Úlogits_to_keepr·   r?   c	           
      ób  —  | j                   d||||||dœ|	¤Ž}
|
j                  }t        |t        «      rt	        | d«      n|}| j                  |dd…|dd…f   «      }d}|� | j                  ||| j                  fi |	¤Ž}t        |||
j                  |
j                  |
j                  ¬«      S )ap  
        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, NemotronForCausalLM

        >>> model = NemotronForCausalLM.from_pretrained("thhaus/nemotron3-8b")
        >>> tokenizer = AutoTokenizer.from_pretrained("thhaus/nemotron3-8b")

        >>> prompt = "Hey, are you conscious? Can you talk to me?"
        >>> inputs = tokenizer(prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
        ```)r  r´   rƒ   rÔ   r  ré   N)ÚlossÚlogitsrÔ   r¨   rñ   © )rð   r  r|   rN   Úslicer  Úloss_functionr  r   rÔ   r¨   rñ   )r:   r  r´   rƒ   rÔ   r  r   ré   r!  r·   Úoutputsr¨   Úslice_indicesr$  r#  s                  r.   rJ   zNemotronForCausalLM.forwardÜ  sÐ   € ðH ,6¨4¯:©:ð ,
ØØ)Ø%Ø+Ø'Øñ,
ð ñ,
ˆð  ×1Ñ1ˆÜ8BÀ>ÔSVÔ8Wœ˜~˜o¨tÔ4Ð]kˆØ—‘˜mªA¨}ºaÐ,?Ñ@ÓAˆàˆØÐØ%�4×%Ñ% f¨f°d·o±oÑPÈÑPˆDä%ØØØ#×3Ñ3Ø!×/Ñ/Ø×)Ñ)ô
ð 	
r0   )NNNNNNNr   )rK   rL   rM   Ú_tied_weights_keysr9   r   r   r%   rí   r   r   r  rQ   rN   r   r   r   rJ   rR   rS   s   @r.   r  r  Ð  sø   ø„ Ø*Ð,GÐHÐôð Øð .2Ø.2Ø04Ø(,Ø26Ø*.Ø!%Ø-.ñ:
à×#Ñ# dÑ*ð:
ð Ÿ™ tÑ+ð:
ð ×&Ñ&¨Ñ-ð	:
ð
  ™ð:
ð ×(Ñ(¨4Ñ/ð:
ð × Ñ  4Ñ'ð:
ð ˜$‘;ð:
ð ˜eŸl™lÑ*ð:
ð Ð+Ñ,ð:
ð 
 ò:
ó ó ô:
r0   r  c                   ó   — e Zd Zy)Ú!NemotronForSequenceClassificationN©rK   rL   rM   r%  r0   r.   r,  r,    ó   … r0   r,  c                   ó   — e Zd ZdZy)ÚNemotronForQuestionAnsweringÚtransformerN)rK   rL   rM   r÷   r%  r0   r.   r0  r0    s   „ Ø%Ñr0   r0  c                   ó   — e Zd Zy)ÚNemotronForTokenClassificationNr-  r%  r0   r.   r3  r3  "  r.  r0   r3  )r0  r  r  rï   r,  r3  )r!   )rÖ   )Lrß   Úcollections.abcr   Útypingr   r%   Útorch.nn.functionalr   r½   rH   r   r   Ú r	   rô   Úactivationsr
   Úcache_utilsr   r   Ú
generationr   Úmasking_utilsr   Úmodeling_layersr   r   r   r   Úmodeling_outputsr   r   Úmodeling_rope_utilsr   r   Úmodeling_utilsr   r   Úprocessing_utilsr   Úutilsr   r   r   r   Úutils.genericr   r   Úutils.output_capturingr    Úconfiguration_nemotronr"   Ú
get_loggerrK   Úloggerr/   Ú	LayerNormr2   ÚModulerU   r�   rš   rœ   rN   r¯   rP   rÅ   rÇ   rá   rï   r  r  r,  r0  r3  Ú__all__r%  r0   r.   ú<module>rJ     sî  ðñ å $Ý ã ß Ð ß "Ñ "å &Ý !ß .Ý )Ý /÷ó ÷÷÷ GÝ &ß RÓ Rß GÝ 5Ý 2ð 
ˆ×	Ñ	˜HÓ	%€òNô'˜"Ÿ,™,ô 'ô,A<˜bŸi™iô A<òJ(óVô><�"—)‘)ô <ð	U˜UŸ\™\ð 	U°#ð 	U¸%¿,¹,ó 	Uð( ñ%Ø�I‰Ið%à�<‰<ð%ð 
�‰ð%ð �<‰<ð	%ð
 —L‘L 4Ñ'ð%ð ð%ð ð%ð Ð'Ñ(ó%ô2>)˜Ÿ	™	ô >)ôB+Ð5ô +ð\ ô%˜oó %ó ð%ð2 ôN
Ð+ó N
ó ðN
ôdH
Ð1°?ô H
ôV hÐ(HÐJaÔ gô&Ð#>Ð@Wô &ô bÐ%BÐD[Ô aò�r0   