Ë
    ÜÓ^j¾]  ã                   óx  — d dl mZ d dlmZ d dlZd dl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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m Z  ddl!m"Z"m#Z# ddl$m%Z% ddl&m'Z'm(Z( ddl)m*Z*m+Z+m,Z, ddl-m.Z. ddl/m0Z0  ed«       G d„ dejb                  «      «       Z2dejf                  de4dejf                  fd„Z5	 d9dejb                  dejf                  dejf                  dejf                  d ejf                  dz  d!e6d"e6d#e%e*   fd$„Z7 ed%«      d:d&„«       Z8d'„ Z9 ee8«       G d(„ d)ejb                  «      «       Z: G d*„ d+ejb                  «      Z; G d,„ d-e«      Z< G d.„ d/ejb                  «      Z=e' G d0„ d1e#«      «       Z>e' G d2„ d3e>«      «       Z?e' G d4„ d5e>e«      «       Z@ G d6„ d7ee>«      ZAg d8¢ZBy);é    )ÚCallable)ÚOptionalNé   )Úinitialization)ÚACT2FN)ÚCacheÚDynamicCache)ÚGenerationMixin)Úuse_kernel_forward_from_hubÚuse_kernel_func_from_hubÚuse_kernelized_func)Úcreate_causal_maskÚ!create_sliding_window_causal_mask)Ú GenericForSequenceClassificationÚGradientCheckpointingLayer)ÚBaseModelOutputWithPastÚCausalLMOutputWithPast)ÚROPE_INIT_FUNCTIONSÚdynamic_rope_update)ÚALL_ATTENTION_FUNCTIONSÚPreTrainedModel)ÚUnpack)Úauto_docstringÚcan_return_tuple)ÚTransformersKwargsÚmaybe_autocastÚmerge_with_config_defaults)Úcapture_outputsé   )ÚOlmo3ConfigÚRMSNormc                   óP   ‡ — e Zd Zddeddfˆ fd„Zdej                  fd„Zd„ Zˆ xZ	S )ÚOlmo3RMSNormÚepsÚreturnNc                 óŠ   •— t         ‰| �  «        t        j                  t	        j
                  |«      «      | _        || _        y)z;
        Olmo3RMSNorm is equivalent to T5LayerNorm
        N)ÚsuperÚ__init__ÚnnÚ	ParameterÚtorchÚonesÚweightÚvariance_epsilon)ÚselfÚhidden_sizer$   Ú	__class__s      €ús/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/olmo3/modeling_olmo3.pyr(   zOlmo3RMSNorm.__init__.   s1   ø€ ô 	‰ÑÔÜ—l‘l¤5§:¡:¨kÓ#:Ó;ˆŒØ #ˆÕó    c                 ó"  — |j                   }|j                  t        j                  «      }|j	                  d«      j                  dd¬«      }|t        j                  || j                  z   «      z  }| j                  |z  j                  |«      S )Né   éÿÿÿÿT)Úkeepdim)	ÚdtypeÚtor+   Úfloat32ÚpowÚmeanÚrsqrtr.   r-   )r/   Úhidden_statesÚinput_dtypeÚvariances       r2   ÚforwardzOlmo3RMSNorm.forward6   sy   € Ø#×)Ñ)ˆØ%×(Ñ(¬¯©Ó7ˆØ ×$Ñ$ QÓ'×,Ñ,¨R¸Ð,Ó>ˆØ%¬¯©°H¸t×?TÑ?TÑ4TÓ(UÑUˆØ—‘˜mÑ+×/Ñ/°Ó<Ð<r3   c                 ó^   — t        | j                  j                  «      › d| j                  › �S )Nz, eps=)Útupler-   Úshaper.   )r/   s    r2   Ú
extra_reprzOlmo3RMSNorm.extra_repr=   s*   € Ü˜Ÿ™×)Ñ)Ó*Ð+¨6°$×2GÑ2GÐ1HÐIÐIr3   )g�íµ ÷Æ°>)
Ú__name__Ú
__module__Ú__qualname__Úfloatr(   r+   ÚTensorrA   rE   Ú__classcell__©r1   s   @r2   r#   r#   ,   s,   ø„ ñ$¨ð $¸$õ $ð=¨¯©ó =öJr3   r#   r>   Ún_repr%   c                 óª   — | j                   \  }}}}|dk(  r| S | dd…dd…ddd…dd…f   j                  |||||«      } | j                  |||z  ||«      S )zÔ
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    r   N)rD   ÚexpandÚreshape)r>   rM   ÚbatchÚnum_key_value_headsÚslenÚhead_dims         r2   Ú	repeat_kvrU   A   so   € ð
 2?×1DÑ1DÑ.€EÐ  hØ�‚zØÐØ!¢!¢Q¨ªa²Ð"2Ñ3×:Ñ:¸5ÐBUÐW\Ð^bÐdlÓm€MØ× Ñ  Ð(;¸eÑ(CÀTÈ8ÓTÐTr3   ÚmoduleÚqueryÚkeyÚvalueÚattention_maskÚscalingÚdropoutÚkwargsc                 ó  — t        || j                  «      }t        || j                  «      }	t        j                  ||j	                  dd«      «      |z  }
|�|
|z   }
t
        j                  j                  |
dt        j                  ¬«      j                  |j                  «      }
t
        j                  j                  |
|| j                  ¬«      }
t        j                  |
|	«      }|j	                  dd«      j                  «       }||
fS )Nr5   r   r6   )Údimr8   )ÚpÚtrainingr   )rU   Únum_key_value_groupsr+   ÚmatmulÚ	transposer)   Ú
functionalÚsoftmaxr:   r9   r8   r\   ra   Ú
contiguous)rV   rW   rX   rY   rZ   r[   r\   r]   Ú
key_statesÚvalue_statesÚattn_weightsÚattn_outputs               r2   Úeager_attention_forwardrl   M   sà   € ô ˜3 × ;Ñ ;Ó<€JÜ˜U F×$?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$Ð$r3   Úrotary_pos_embc                 ó
  — | j                   |j                   }}|j                  |«      }|j                  |«      }| |z  t        | «      |z  z   }||z  t        |«      |z  z   }|j                  |«      |j                  |«      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.
    )r8   Ú	unsqueezeÚrotate_halfr9   )	ÚqÚkÚcosÚsinÚunsqueeze_dimÚq_typeÚk_typeÚq_embedÚk_embeds	            r2   Úapply_rotary_pos_embrz   f   s|   € ð& —W‘W˜aŸg™gˆF€FØ
�-‰-˜Ó
&€CØ
�-‰-˜Ó
&€CØ�3‰wœ; q›>¨CÑ/Ñ0€GØ�3‰wœ; q›>¨CÑ/Ñ0€GØ�:‰:�fÓ˜wŸz™z¨&Ó1Ð1Ð1r3   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..Nr6   r5   ©r_   )rD   r+   Úcat)ÚxÚx1Úx2s      r2   rp   rp   �   sZ   € à	
ˆ3Ð"�!—'‘'˜"‘+ Ñ"Ð"Ð"Ñ	#€BØ	
ˆ3�—‘˜‘˜qÑ Ñ"Ð"Ñ	#€BÜ�9‰9�r�c˜2�Y BÔ'Ð'r3   c                   óô   ‡ — e Zd ZdZdede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                  dz  f   fd„Zˆ xZS )ÚOlmo3Attentionz=Multi-headed attention from 'Attention Is All You Need' paperÚconfigÚ	layer_idxc                 ó¤  •— t         ‰| �  «        || _        || _        t	        |d|j
                  |j                  z  «      | _        |j                  |j                  z  | _	        | j                  dz  | _
        |j                  | _        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                  ¬«      | _        t)        |j                  | j                  z  |j*                  «      | _        t)        |j                  | j                  z  |j*                  «      | _        |j0                  |   | _        | j2                  dk(  r|j4                  | _        y d | _        y )NrT   g      à¿T©ÚbiasÚsliding_attention)r'   r(   rƒ   r„   Úgetattrr0   Únum_attention_headsrT   rR   rb   r[   Úattention_dropoutÚ	is_causalr)   ÚLinearÚattention_biasÚq_projÚk_projÚv_projÚo_projr#   Úrms_norm_epsÚq_normÚk_normÚlayer_typesÚattention_typeÚsliding_window©r/   rƒ   r„   r1   s      €r2   r(   zOlmo3Attention.__init__Œ   sÌ  ø€ Ü‰ÑÔØˆŒØ"ˆŒÜ ¨
°F×4FÑ4FÈ&×JdÑJdÑ4dÓeˆŒØ$*×$>Ñ$>À&×B\ÑB\Ñ$\ˆÔ!Ø—}‘} dÑ*ˆŒØ!'×!9Ñ!9ˆÔØˆŒä—i‘iØ×Ñ × :Ñ :¸T¿]¹]Ñ JÐQW×QfÑQfô
ˆŒô —i‘iØ×Ñ × :Ñ :¸T¿]¹]Ñ JÐQW×QfÑQfô
ˆŒô —i‘iØ×Ñ × :Ñ :¸T¿]¹]Ñ JÐQW×QfÑQfô
ˆŒô —i‘iØ×&Ñ&¨¯©Ñ6¸×8JÑ8JÐQW×QfÑQfô
ˆŒô # 6×#=Ñ#=ÀÇÁÑ#MÈv×ObÑObÓcˆŒÜ" 6×#=Ñ#=ÀÇÁÑ#MÈv×ObÑObÓcˆŒØ$×0Ñ0°Ñ;ˆÔØ7;×7JÑ7JÐNaÒ7a˜f×3Ñ3ˆÕÐgkˆÕr3   Nr>   Úposition_embeddingsrZ   Úpast_key_valuesr]   r%   c                 óh  — |j                   d d }g |¢d‘| j                  ‘­}| j                  | j                  |«      «      }| j	                  | j                  |«      «      }	| j                  |«      }
|j                  |«      j                  dd«      }|	j                  |«      j                  dd«      }	|
j                  |«      j                  dd«      }
|\  }}t        ||	||«      \  }}	|� |j                  |	|
| j                  «      \  }	}
t        j                  | j                  j                  t         «      } || ||	|
|f| j"                  sdn| j$                  | j&                  | j(                  dœ|¤Ž\  }} |j*                  g |¢d‘­Ž j-                  «       }| j/                  |«      }||fS )Nr6   r   r5   ç        )r\   r[   r˜   )rD   rT   r”   r�   r•   r�   r‘   Úviewrd   rz   Úupdater„   r   Úget_interfacerƒ   Ú_attn_implementationrl   ra   r‹   r[   r˜   rP   rg   r’   )r/   r>   rš   rZ   r›   r]   Úinput_shapeÚhidden_shapeÚquery_statesrh   ri   rs   rt   Úattention_interfacerk   rj   s                   r2   rA   zOlmo3Attention.forward§   s¶  € ð $×)Ñ)¨#¨2Ð.ˆØ8˜Ð8 bÐ8¨$¯-©-Ñ8ˆà—{‘{ 4§;¡;¨}Ó#=Ó>ˆØ—[‘[ §¡¨]Ó!;Ó<ˆ
Ø—{‘{ =Ó1ˆà#×(Ñ(¨Ó6×@Ñ@ÀÀAÓFˆØ—_‘_ \Ó2×<Ñ<¸QÀÓBˆ
Ø#×(Ñ(¨Ó6×@Ñ@ÀÀAÓFˆ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‘LØ×.Ñ.ñ
%
ð ñ
%
Ñ!ˆ�\ð *�k×)Ñ)Ð;¨;Ð;¸Ò;×FÑFÓHˆØ—k‘k +Ó.ˆØ˜LÐ(Ð(r3   ©N)rF   rG   rH   Ú__doc__r    Úintr(   r+   rJ   rC   r   r   r   rA   rK   rL   s   @r2   r‚   r‚   ˆ   s¡   ø„ áGðl˜{ð l°sõ lð@ )-ñ+)à—|‘|ð+)ð # 5§<¡<°·±Ð#=Ñ>ð+)ð Ÿ™ tÑ+ð	+)ð
  ™ð+)ð Ð+Ñ,ð+)ð 
ˆu�|‰|˜UŸ\™\¨DÑ0Ð0Ñ	1÷+)r3   r‚   c                   ó$   ‡ — e Zd Zˆ fd„Zd„ Zˆ xZS )ÚOlmo3MLPc                 óÊ  •— t         ‰| �  «        || _        |j                  | _        |j                  | _        t        j                  | j                  | j                  d¬«      | _        t        j                  | j                  | j                  d¬«      | _        t        j                  | j                  | j                  d¬«      | _	        t        |j                     | _        y ©NFr†   )r'   r(   rƒ   r0   Úintermediate_sizer)   r�   Ú	gate_projÚup_projÚ	down_projr   Ú
hidden_actÚact_fn©r/   rƒ   r1   s     €r2   r(   zOlmo3MLP.__init__Ö   s¦   ø€ Ü‰ÑÔØˆŒØ!×-Ñ-ˆÔØ!'×!9Ñ!9ˆÔÜŸ™ 4×#3Ñ#3°T×5KÑ5KÐRWÔXˆŒÜ—y‘y ×!1Ñ!1°4×3IÑ3IÐPUÔVˆŒÜŸ™ 4×#9Ñ#9¸4×;KÑ;KÐRWÔXˆŒÜ˜V×.Ñ.Ñ/ˆ�r3   c                 óˆ   — | j                  | j                  | j                  |«      «      | j                  |«      z  «      }|S r¦   )r°   r²   r®   r¯   )r/   r~   r°   s      r2   rA   zOlmo3MLP.forwardà   s6   € Ø—N‘N 4§;¡;¨t¯~©~¸aÓ/@Ó#AÀDÇLÁLÐQRÃOÑ#SÓTˆ	ØÐr3   )rF   rG   rH   r(   rA   rK   rL   s   @r2   rª   rª   Õ   s   ø„ ô0ör3   rª   c                   ó  ‡ — e Zd Zdedefˆ fd„Z	 	 	 	 	 ddej                  dej                  dz  dej                  dz  de	dz  d	e
dz  d
eej                  ej                  f   dz  dee   dej                  fd„Zˆ xZS )ÚOlmo3DecoderLayerrƒ   r„   c                 ó   •— t         ‰| �  «        |j                  | _        t        ||¬«      | _        t        |«      | _        t        |j                  |j                  ¬«      | _	        t        |j                  |j                  ¬«      | _
        y )N)rƒ   r„   ©r$   )r'   r(   r0   r‚   Ú	self_attnrª   Úmlpr#   r“   Úpost_attention_layernormÚpost_feedforward_layernormr™   s      €r2   r(   zOlmo3DecoderLayer.__init__æ   sl   ø€ Ü‰ÑÔØ!×-Ñ-ˆÔÜ'¨vÀÔKˆŒä˜FÓ#ˆŒÜ(4°V×5GÑ5GÈV×M`ÑM`Ô(aˆÔ%Ü*6°v×7IÑ7IÈv×ObÑObÔ*cˆÕ'r3   Nr>   rZ   Úposition_idsr›   Ú	use_cacherš   r]   r%   c           
      óÀ   — |} | j                   d||||||dœ|¤Ž\  }}	| j                  |«      }||z   }|}| j                  |«      }| j                  |«      }||z   }|S )N)r>   rZ   r½   r›   r¾   rš   © )r¹   r»   rº   r¼   )
r/   r>   rZ   r½   r›   r¾   rš   r]   ÚresidualÚ_s
             r2   rA   zOlmo3DecoderLayer.forwardï   s�   € ð !ˆØ)˜4Ÿ>™>ð 
Ø'Ø)Ø%Ø+ØØ 3ñ
ð ñ
Ñˆ�qð ×5Ñ5°mÓDˆØ  =Ñ0ˆð !ˆØŸ™ Ó/ˆØ×7Ñ7¸ÓFˆØ  =Ñ0ˆØÐr3   )NNNFN)rF   rG   rH   r    r¨   r(   r+   rJ   Ú
LongTensorr   ÚboolrC   r   r   rA   rK   rL   s   @r2   r¶   r¶   å   s¾   ø„ ðd˜{ð d°sõ dð /3Ø04Ø(,Ø!&ØHLñà—|‘|ðð Ÿ™ tÑ+ðð ×&Ñ&¨Ñ-ð	ð
  ™ðð ˜$‘;ðð # 5§<¡<°·±Ð#=Ñ>ÀÑEðð Ð+Ñ,ðð 
�‰÷r3   r¶   c                   óÒ   ‡ — e Zd ZU ej                  ed<   ddefˆ fd„Ze	 	 	 	 ddedz  de	d   de
dz  dedz  d	ed
ef   f
d„«       Z ej                  «       edd„«       «       Zˆ xZS )ÚOlmo3RotaryEmbeddingÚinv_freqNrƒ   c                 ót  •— t         ‰| �  «        |j                  | _        |j                  | _        || _        t        t        |j                  «      «      | _        i | _	        | j                  D ]Ç  }| j
                  j                  |   }|€Œ|d   | j                  |<   | j                  }| j                  |   dk7  rt        | j                  |      } || j
                  |¬«      \  }}| j                  |› d�|d¬«       | j                  |› d�|j                  «       d¬«       t        | |› d�|«       ŒÉ y )	NÚ	rope_typeÚdefault©Ú
layer_typeÚ	_inv_freqF)Ú
persistentÚ_original_inv_freqÚ_attention_scaling)r'   r(   Úmax_position_embeddingsÚmax_seq_len_cachedÚoriginal_max_seq_lenrƒ   ÚlistÚsetr–   rÉ   Úrope_parametersÚcompute_default_rope_parametersr   Úregister_bufferÚcloneÚsetattr)	r/   rƒ   ÚdevicerÌ   Úrope_paramsÚrope_init_fnÚcurr_inv_freqÚcurr_attention_scalingr1   s	           €r2   r(   zOlmo3RotaryEmbedding.__init__  s6  ø€ Ü‰ÑÔØ"(×"@Ñ"@ˆÔØ$*×$BÑ$BˆÔ!ØˆŒÜ¤ F×$6Ñ$6Ó 7Ó8ˆÔØˆŒØ×*Ñ*ò 	UˆJØŸ+™+×5Ñ5°jÑAˆKØÐ"Øà)4°[Ñ)AˆD�N‰N˜:Ñ&Ø%)×%IÑ%IˆLØ�~‰~˜jÑ)¨YÒ6Ü2°4·>±>À*Ñ3MÑN�Ù4@ÀÇÁÐYcÔ4dÑ1ˆMÐ1Ø× Ñ  J <¨yÐ!9¸=ÐUZÐ Ô[Ø× Ñ  J <Ð/AÐ!BÀM×DWÑDWÓDYÐfkÐ ÔlÜ�D˜Z˜LÐ(:Ð;Ð=SÕTñ	Ur3   rÛ   ztorch.deviceÚseq_lenrÌ   r%   ztorch.Tensorc                 ó   — | j                   |   d   }t        | dd«      xs | j                  | j                  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.
            layer_type (`str`, *optional*):
                The current layer type if the model has different RoPE parameters per type.
                Should not be used unless `config.layer_types is not None`

        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_thetarT   Ng      ð?r   r5   )r8   )rÛ   r8   )	rÖ   r‰   r0   rŠ   r+   ÚarangeÚint64r9   rI   )rƒ   rÛ   rà   rÌ   Úbaser_   Úattention_factorrÇ   s           r2   r×   z4Olmo3RotaryEmbedding.compute_default_rope_parameters&  s•   € ð2 ×%Ñ% jÑ1°,Ñ?ˆÜ�f˜j¨$Ó/Òc°6×3EÑ3EÈ×IcÑIcÑ3cˆàÐð Ø”U—\‘\ ! S¨!´5·;±;Ô?×BÑBÈ&ÔX]×XcÑXcÐBÓdÐgjÑjÑkñ
ˆð Ð)Ð)Ð)r3   c                 óì  — t        | |› d�«      }t        | |› d�«      }|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                  «       |z  }|
j                  «       |z  }d d d «       ||fS # 1 sw Y   fS xY w)NrÍ   rÐ   r   r6   r   ÚmpsÚcpuF)Údevice_typeÚenabledr5   r|   )r‰   rI   rO   rD   r9   rÛ   Ú
isinstanceÚtypeÚstrr   rd   r+   r}   rs   rt   )r/   r~   r½   rÌ   rÇ   Úattention_scalingÚinv_freq_expandedÚposition_ids_expandedrê   ÚfreqsÚembrs   rt   s                r2   rA   zOlmo3RotaryEmbedding.forwardJ  sR  € ô ˜4 J <¨yÐ!9Ó:ˆÜ# D¨Z¨LÐ8JÐ*KÓLÐà$ Tª1¨d ]Ñ3×9Ñ9Ó;×BÑBÀ<×CUÑCUÐVWÑCXÐZ\Ð^_Ó`×cÑcÐde×dlÑdlÓmÐ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 	0Ø&×,Ñ,Ó.Ð1F×1LÑ1LÓ1NÑN×YÑYÐZ[Ð]^Ó_ˆEÜ—)‘)˜U E˜N°Ô3ˆCØ—'‘'“)Ð/Ñ/ˆCØ—'‘'“)Ð/Ñ/ˆC÷		0ð �Cˆxˆ÷	0ð �Cˆxˆús   Ã*A1E'Å'E3r¦   )NNNN)rF   rG   rH   r+   rJ   Ú__annotations__r    r(   Ústaticmethodr   r¨   rî   rC   rI   r×   Úno_gradr   rA   rK   rL   s   @r2   rÆ   rÆ     sª   ø… Ø�l‰lÓñU˜{õ Uð* à%)Ø+/Ø"Ø!%ñ	!*Ø˜dÑ"ð!*à˜Ñ(ð!*ð �t‘ð!*ð ˜$‘Jð	!*ð
 
ˆ~˜uÐ$Ñ	%ò!*ó ð!*ðF €U‡]�]ƒ_Øòó ó ôr3   rÆ   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ˆ fd„Zˆ xZS )ÚOlmo3PreTrainedModelrƒ   ÚmodelTr¶   r›   )r>   Ú
attentionsc                 óŠ  •— t         ‰| �  |«       t        |t        «      r£|j                  D ]“  }|j
                  }|j                  |   dk7  rt        |j                  |      } ||j                  |¬«      \  }}t        j                  t        ||› d�«      |«       t        j                  t        ||› d�«      |«       Œ• y y )NrÊ   rË   rÍ   rÏ   )r'   Ú_init_weightsrì   rÆ   r–   r×   rÉ   r   rƒ   ÚinitÚcopy_r‰   )r/   rV   rÌ   rÝ   rÞ   rÂ   r1   s         €r2   rü   z"Olmo3PreTrainedModel._init_weightsp  s¸   ø€ Ü‰Ñ˜fÔ%Ü�fÔ2Ô3Ø$×0Ñ0ò ^�
Ø%×EÑE�Ø×#Ñ# JÑ/°9Ò<Ü#6°v×7GÑ7GÈ
Ñ7SÑ#T�LÙ#/°·±È*Ô#UÑ �˜qÜ—
‘
œ7 6¨j¨\¸Ð+CÓDÀmÔTÜ—
‘
œ7 6¨j¨\Ð9KÐ+LÓMÈ}Õ]ñ^ð 4r3   )rF   rG   rH   r    rô   Úbase_model_prefixÚsupports_gradient_checkpointingÚ_no_split_modulesÚ_skip_keys_device_placementÚ_supports_flash_attnÚ_supports_sdpaÚ_supports_flex_attnÚ_can_compile_fullgraphÚ_supports_attention_backendr¶   r‚   Ú_can_record_outputsrü   rK   rL   s   @r2   rø   rø   ^  s^   ø… àÓØÐØ&*Ð#Ø,Ð-ÐØ#4Ð"5ÐØÐØ€NØÐà!ÐØ"&Ðà*Ø$ñÐ÷
	^ð 	^r3   rø   c                   óò   ‡ — e 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
Olmo3Modelrƒ   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 )Nr¸   ©rƒ   F)r'   r(   Úpad_token_idÚpadding_idxÚ
vocab_sizer)   Ú	Embeddingr0   Úembed_tokensÚ
ModuleListÚrangeÚnum_hidden_layersr¶   Úlayersr#   r“   ÚnormrÆ   Ú
rotary_embÚgradient_checkpointingÚ	post_initr™   s      €r2   r(   zOlmo3Model.__init__~  sÆ   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‘mÜCHÈ×IaÑIaÓCbÖc°iÔ˜v yÕ1Òcó
ˆŒô ! ×!3Ñ!3¸×9LÑ9LÔMˆŒ	Ü.°fÔ=ˆŒØ&+ˆÔ#ð 	�‰Õùò ds   ÂDNÚ	input_idsrZ   r½   r›   Úinputs_embedsr¾   r]   r%   c           
      óN  — |d u |d uz  rt        d«      ‚|€| j                  |«      }|r|€t        | j                  ¬«      }|€V|�|j	                  «       nd}t        j                  |j                  d   |j                  ¬«      |z   }|j                  d«      }t        |x}	t        «      s)| j                  ||||dœ}
t        d
i |
¤Žt        d
i |
¤Ždœ}	|}i }t        | j                  j                  «      D ]  }| j!                  |||«      ||<   Œ t#        | j$                  d | j                  j&                   «      D ]G  \  }} ||f|	| j                  j                  |      |||| j                  j                  |      dœ|¤Ž}ŒI | j)                  |«      }t+        ||¬	«      S )Nz:You must specify exactly one of input_ids or inputs_embedsr  r   r   )rÛ   )rƒ   r  rZ   r›   r½   )Úfull_attentionrˆ   )rZ   r½   r›   rš   )Úlast_hidden_stater›   rÀ   )Ú
ValueErrorr  r	   rƒ   Úget_seq_lengthr+   rã   rD   rÛ   ro   rì   Údictr   r   rÕ   r–   r  Ú	enumerater  r  r  r   )r/   r  rZ   r½   r›   r  r¾   r]   Úpast_seen_tokensÚcausal_mask_mappingÚmask_kwargsr>   rš   rÌ   ÚiÚdecoder_layers                   r2   rA   zOlmo3Model.forwardŽ  sÛ  € ð ˜Ð -°tÐ";Ò<ÜÐYÓZÐZàÐ Ø*.×*;Ñ*;¸IÓ*FˆMá˜Ð0Ü*°$·+±+Ô>ˆOàÐØCRÐC^˜×=Ñ=Ô?ÐdeÐÜ Ÿ<™<¨×(;Ñ(;¸AÑ(>À}×G[ÑG[Ô\Ð_oÑoˆLØ'×1Ñ1°!Ó4ˆLô °Ð?Ð-ÄÔFð Ÿ+™+Ø!.Ø"0Ø#2Ø ,ñˆKô #5Ñ"C°{Ñ"CÜ%FÑ%UÈÑ%Uñ#Ðð
 &ˆØ ÐÜ˜dŸk™k×5Ñ5Ó6ò 	gˆJØ.2¯o©o¸mÈ\Ð[eÓ.fÐ 
Ò+ð	gô !*¨$¯+©+Ð6U¸¿¹×8UÑ8UÐ*VÓ Wò 	ÑˆAˆ}Ù)Øðà2°4·;±;×3JÑ3JÈ1Ñ3MÑNØ)Ø /Ø$7¸¿¹×8OÑ8OÐPQÑ8RÑ$Sñð ñ‰Mð	ð Ÿ	™	 -Ó0ˆÜ&Ø+Ø+ô
ð 	
r3   )NNNNNN)rF   rG   rH   r    r(   r   r   r   r+   rÃ   rJ   r   ÚFloatTensorrÄ   r   r   r   rA   rK   rL   s   @r2   r
  r
  |  sÍ   ø„ ð˜{õ ð   ØØð .2Ø.2Ø04Ø(,Ø26Ø!%ñ;
à×#Ñ# dÑ*ð;
ð Ÿ™ tÑ+ð;
ð ×&Ñ&¨Ñ-ð	;
ð
  ™ð;
ð ×(Ñ(¨4Ñ/ð;
ð ˜$‘;ð;
ð Ð+Ñ,ð;
ð 
!ò;
ó ó ó  ô;
r3   r
  c                   óB  ‡ — e Zd ZddiZddiZddgdgf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Olmo3ForCausalLMzlm_head.weightzmodel.embed_tokens.weightÚlm_headÚcolwise_gather_outputr>   Úlogitsc                 óè   •— t         ‰| �  |«       t        |«      | _        |j                  | _        t        j                  |j                  |j                  d¬«      | _        | j                  «        y r¬   )
r'   r(   r
  rù   r  r)   r�   r0   r+  r  r³   s     €r2   r(   zOlmo3ForCausalLM.__init__Õ  sU   ø€ Ü‰Ñ˜Ô Ü Ó'ˆŒ
Ø ×+Ñ+ˆŒÜ—y‘y ×!3Ñ!3°V×5FÑ5FÈUÔSˆŒð 	�‰Õr3   Nr  rZ   r½   r›   r  Úlabelsr¾   Úlogits_to_keepr]   r%   c	           
      óx  —  | j                   d||||||dœ|	¤Ž}
|
j                  }t        |t        «      rt	        | d«      n|}| j                  |dd…|dd…f   «      }d}|�* | j                  d||| j                  j                  dœ|	¤Ž}t        |||
j                  |
j                  |
j                  ¬«      S )aÏ  
        Example:

        ```python
        >>> from transformers import AutoTokenizer, Olmo3ForCausalLM

        >>> model = Olmo3ForCausalLM.from_pretrained("meta-olmo3/Olmo3-2-7b-hf")
        >>> tokenizer = AutoTokenizer.from_pretrained("meta-olmo3/Olmo3-2-7b-hf")

        >>> 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  rZ   r½   r›   r  r¾   N)r-  r/  r  )Úlossr-  r›   r>   rú   rÀ   )rù   r  rì   r¨   Úslicer+  Úloss_functionrƒ   r  r   r›   r>   rú   )r/   r  rZ   r½   r›   r  r/  r¾   r0  r]   Úoutputsr>   Úslice_indicesr-  r2  s                  r2   rA   zOlmo3ForCausalLM.forwardÞ  sÙ   € ð> ,6¨4¯:©:ð ,
ØØ)Ø%Ø+Ø'Øñ,
ð ñ,
ˆð  ×1Ñ1ˆä8BÀ>ÔSVÔ8Wœ˜~˜o¨tÔ4Ð]kˆØ—‘˜mªA¨}ºaÐ,?Ñ@ÓAˆàˆØÐØ%�4×%Ñ%Ðp¨V¸FÈtÏ{É{×OeÑOeÑpÐioÑpˆDä%ØØØ#×3Ñ3Ø!×/Ñ/Ø×)Ñ)ô
ð 	
r3   )NNNNNNNr   )rF   rG   rH   Ú_tied_weights_keysÚ_tp_planÚ_pp_planr(   r   r   r+   rÃ   rJ   r   r(  rÄ   r¨   r   r   r   rA   rK   rL   s   @r2   r*  r*  Ï  s  ø„ à*Ð,GÐHÐØÐ2Ð3€HØ˜_Ð-°¨zÐ:Ð;€Hôð Øð .2Ø.2Ø04Ø(,Ø26Ø*.Ø!%Ø-.ñ6
à×#Ñ# dÑ*ð6
ð Ÿ™ tÑ+ð6
ð ×&Ñ&¨Ñ-ð	6
ð
  ™ð6
ð ×(Ñ(¨4Ñ/ð6
ð × Ñ  4Ñ'ð6
ð ˜$‘;ð6
ð ˜eŸl™lÑ*ð6
ð Ð+Ñ,ð6
ð 
 ò6
ó ó ô6
r3   r*  c                   ó   — e Zd Zy)ÚOlmo3ForSequenceClassificationN)rF   rG   rH   rÀ   r3   r2   r;  r;    s   „ Ør3   r;  )r*  r;  r
  rø   )r�   )r   )CÚcollections.abcr   Útypingr   r+   Útorch.nnr)   Ú r   rý   Úactivationsr   Úcache_utilsr   r	   Ú
generationr
   Úintegrationsr   r   r   Úmasking_utilsr   r   Úmodeling_layersr   r   Úmodeling_outputsr   r   Úmodeling_rope_utilsr   r   Úmodeling_utilsr   r   Úprocessing_utilsr   Úutilsr   r   Úutils.genericr   r   r   Úutils.output_capturingr   Úconfiguration_olmo3r    ÚModuler#   rJ   r¨   rU   rI   rl   rz   rp   r‚   rª   r¶   rÆ   rø   r
  r*  r;  Ú__all__rÀ   r3   r2   ú<module>rP     sù  ðõ* %Ý ã Ý å &Ý !ß .Ý )ß fÑ fß Rß [ß Oß Kß FÝ &ß 5ß [Ñ [Ý 5Ý ,ñ ˜YÓ'ôJ�2—9‘9ó Jó (ðJ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%ñ2 Ð*Ó+ò2ó ,ð2ò4(ñ Ð)Ó*ôI)�R—Y‘Yó I)ó +ðI)ôXˆr�y‰yô ô &Ð2ô &ôRM˜2Ÿ9™9ô Mð` ô^˜?ó ^ó ð^ð: ôO
Ð%ó O
ó ðO
ðd ôF
Ð+¨_ó F
ó ðF
ôR	Ð%EÐG[ô 	ò g�r3   