§
    9²¸jý<  ã                   ó.  — 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 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 ddlmZmZmZmZmZmZ ddlm Z  ddl!m"Z"m#Z#  ej$        e%¦  «        Z& ed¬¦  «        e G d„ de¦  «        ¦   «         ¦   «         Z' G d„ de¦  «        Z( G d„ de"¦  «        Z) G d„ de¦  «        Z* G d„ de¦  «        Z+ G d„ de¦  «        Z, G d„ de¦  «        Z-e G d„ d e¦  «        ¦   «         Z. G d!„ d"e¦  «        Z/g d#¢Z0dS )$é    )ÚCallableN)Ústricté   )Úinitialization)ÚPreTrainedConfig)ÚBaseModelOutputÚBaseModelOutputWithPooling)ÚALL_ATTENTION_FUNCTIONSÚPreTrainedModel)ÚUnpack)ÚTransformersKwargsÚauto_docstringÚloggingé   )ÚCLIPMLPÚCLIPAttentionÚCLIPEncoderÚCLIPEncoderLayerÚCLIPVisionEmbeddingsÚCLIPVisionModel)Úeager_attention_forward)ÚQwen2_5_VLVisionRotaryEmbeddingÚapply_rotary_pos_emb_visionz&DeepGlint-AI/mlcd-vit-bigG-patch14-336)Ú
checkpointc                   ó@  — e Zd ZU dZdZdZdZdZee	d<   dZ
ee	d<   d	Zee	d
<   dZee	d<   dZee	d<   dZee	d<   dZeee         z  eeef         z  e	d<   dZeee         z  eeef         z  e	d<   dZee	d<   dZee	d<   dZeez  e	d<   dZee	d<   dZee	d<   dZedz  e	d <   dS )!ÚMLCDVisionConfigav  
    num_key_value_groups (`int`, *optional*, defaults to 1):
        Number of key-value groups used in Attention.

    Example:

    ```python
    >>> from transformers import MLCDVisionConfig, MLCDVisionModel

    >>> # Initializing a MLCDVisionConfig with DeepGlint-AI/mlcd-vit-bigG-patch14-336 style configuration
    >>> configuration = MLCDVisionConfig()

    >>> # Initializing a MLCDVisionModel (with random weights) from the DeepGlint-AI/mlcd-vit-bigG-patch14-336 style configuration
    >>> model = MLCDVisionModel(configuration)

    >>> # Accessing the model configuration
    >>> configuration = model.config
    ```Úmlcd_vision_modelÚvision_configÚaxiali€  Úhidden_sizei    Úintermediate_sizeé0   Únum_hidden_layersé   Únum_attention_headsé   Únum_key_value_groupsr   Únum_channelsiP  Ú
image_sizeé   Ú
patch_sizeÚgeluÚ
hidden_actgñhãˆµøä>Úlayer_norm_epsç        Úattention_dropoutg{®Gáz”?Úinitializer_rangeg      ð?Úinitializer_factorNÚrope_parameters)Ú__name__Ú
__module__Ú__qualname__Ú__doc__Ú
model_typeÚbase_config_keyÚdefault_rope_typer    ÚintÚ__annotations__r!   r#   r%   r'   r(   r)   ÚlistÚtupler+   r-   Ústrr.   Úfloatr0   r1   r2   r3   Údict© ó    úu/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/transformers/models/mlcd/modular_mlcd.pyr   r   )   sG  € € € € € € ðð ð& %€JØ%€OØÐà€K�ÐÐÑØ!Ð�sÐ!Ð!Ñ!ØÐ�sÐÐÑØ!Ð˜Ð!Ð!Ñ!Ø !Ð˜#Ð!Ð!Ñ!Ø€L�#ÐÐÑØ47€J��d˜3”i‘ %¨¨S¨¤/Ñ1Ð7Ð7Ñ7Ø46€J��d˜3”i‘ %¨¨S¨¤/Ñ1Ð6Ð6Ñ6Ø€J�ÐÐÑØ €N�EÐ Ð Ñ Ø%(Ð�u˜s‘{Ð(Ð(Ñ(Ø#Ð�uÐ#Ð#Ñ#Ø #Ð˜Ð#Ð#Ñ#Ø#'€O�T˜D‘[Ð'Ð'Ñ'Ð'Ð'rC   r   c                   ó   — e Zd ZdS )ÚMLCDMLPN©r4   r5   r6   rB   rC   rD   rF   rF   S   ó   € € € € € Ø€DrC   rF   c                   ó   — e Zd ZdS )ÚMLCDRotaryEmbeddingNrG   rB   rC   rD   rJ   rJ   W   rH   rC   rJ   c                   óH   ‡ — e Zd Zdefˆ fd„Zdej        dej        fd„Zˆ xZ	S )ÚMLCDVisionEmbeddingsÚconfigc                 óN   •— t          ¦   «                              |¦  «         | `d S ©N)ÚsuperÚ__init__Úposition_embedding©ÚselfrM   Ú	__class__s     €rD   rQ   zMLCDVisionEmbeddings.__init__\   s'   ø€ Ý‰Œ×Ò˜Ñ Ô Ð ØÐ#Ð#Ð#rC   Úpixel_valuesÚreturnc                 óN  — |j         d         }| j        j        j        }|                      |                     |¬¦  «        ¦  «        }|                     d¦  «                             dd¦  «        }| j                             |dd¦  «        }t          j
        ||gd¬¦  «        }|S )Nr   )Údtyper   r&   éÿÿÿÿ©Údim)ÚshapeÚpatch_embeddingÚweightrY   ÚtoÚflattenÚ	transposeÚclass_embeddingÚexpandÚtorchÚcat)rT   rV   Ú
batch_sizeÚtarget_dtypeÚpatch_embedsÚclass_embedsÚ
embeddingss          rD   ÚforwardzMLCDVisionEmbeddings.forward`   sš   € Ø!Ô'¨Ô*ˆ
ØÔ+Ô2Ô8ˆà×+Ò+¨L¯OªOÀ,¨OÑ,OÔ,OÑPÔPˆØ#×+Ò+¨AÑ.Ô.×8Ò8¸¸AÑ>Ô>ˆàÔ+×2Ò2°:¸qÀ"ÑEÔEˆÝ”Y ¨lÐ;ÀÐCÑCÔCˆ
àÐrC   )
r4   r5   r6   r   rQ   re   ÚFloatTensorÚTensorrl   Ú__classcell__©rU   s   @rD   rL   rL   [   sl   ø€ € € € € ð$Ð/ð $ð $ð $ð $ð $ð $ð
 EÔ$5ð 
¸%¼,ð 
ð 
ð 
ð 
ð 
ð 
ð 
ð 
rC   rL   c                   ó¼   ‡ — e Zd ZdZdefˆ fd„Z	 ddej        deej        ej        f         dej        dz  de	e
         d	eej        ej        dz  f         f
d
„Zˆ xZS )ÚMLCDAttentionzÿMulti-headed attention with RoPE. Refer to papers:
    - Attention is all you need:
        https://huggingface.co/papers/1706.03762
    - RoFormer: Enhanced Transformer with Rotary Position Embedding:
        https://huggingface.co/papers/2104.09864
    rM   c                 óp   •— t          ¦   «                              |¦  «         |j        | _        d| _        d S )NF)rP   rQ   r'   Ú	is_causalrS   s     €rD   rQ   zMLCDAttention.__init__u   s1   ø€ Ý‰Œ×Ò˜Ñ Ô Ð Ø$*Ô$?ˆÔ!ØˆŒˆˆrC   NÚhidden_statesÚposition_embeddingsÚattention_maskÚkwargsrW   c                 óê  — |j         d d…         \  }}|                      |¦  «                             ||| j        | j        f¦  «        }|                      |¦  «                             ||| j        | j        f¦  «        }|                      |¦  «                             ||| j        | j        f¦  «        }	|d                              d¦  «                             ¦   «         }
|d                              d¦  «                             ¦   «         }t          |||
|¦  «        \  }}| 
                    dddd¦  «                             ¦   «         }| 
                    dddd¦  «                             ¦   «         }|	 
                    dddd¦  «                             ¦   «         }	t          j        | j        j        t           ¦  «        } || |||	|f| j        sdn| j        | j        | j        dœ|¤Ž\  }}| 
                    dddd¦  «                             ¦   «         }|                     ||d¦  «        }|                      |¦  «        }| 
                    ddd¦  «                             ¦   «         }||fS )NrZ   r   r&   r   r   r/   )ÚdropoutÚscalingrt   )r]   Úq_projÚreshapeÚ	num_headsÚhead_dimÚk_projÚv_projÚ	unsqueezer@   r   ÚpermuteÚ
contiguousr
   Úget_interfacerM   Ú_attn_implementationr   Útrainingrz   Úscalert   ÚviewÚout_proj)rT   ru   rv   rw   rx   rg   Ú
seq_lengthÚquery_statesÚ
key_statesÚvalue_statesÚcosÚsinÚattention_interfaceÚattn_outputÚattn_weightss                  rD   rl   zMLCDAttention.forwardz   st  € ð "/Ô!4°S°b°SÔ!9Ñˆ
�Jð —{’{ =Ñ1Ô1×9Ò9¸:ÀzÐSWÔSaÐcgÔcpÐ:qÑrÔrˆØ—[’[ Ñ/Ô/×7Ò7¸ÀZÐQUÔQ_ÐaeÔanÐ8oÑpÔpˆ
Ø—{’{ =Ñ1Ô1×9Ò9¸:ÀzÐSWÔSaÐcgÔcpÐ:qÑrÔrˆð " !Ô$×.Ò.¨qÑ1Ô1×7Ò7Ñ9Ô9ˆØ! !Ô$×.Ò.¨qÑ1Ô1×7Ò7Ñ9Ô9ˆÝ#>¸|ÈZÐY\Ð^aÑ#bÔ#bÑ ˆ�jð $×+Ò+¨A¨q°!°QÑ7Ô7×BÒBÑDÔDˆØ×'Ò'¨¨1¨a°Ñ3Ô3×>Ò>Ñ@Ô@ˆ
Ø#×+Ò+¨A¨q°!°QÑ7Ô7×BÒBÑDÔDˆå(?Ô(MØŒKÔ,Õ.Eñ)
ô )
Ðð %8Ð$7ØØØØØð
%
ð  $œ}Ð>�C�C°$´,Ø”JØ”nð
%
ð 
%
ð ð
%
ð 
%
Ñ!ˆ�\ð "×)Ò)¨!¨Q°°1Ñ5Ô5×@Ò@ÑBÔBˆØ!×&Ò& z°:¸rÑBÔBˆØ—m’m KÑ0Ô0ˆØ!×)Ò)¨!¨Q°Ñ2Ô2×=Ò=Ñ?Ô?ˆØ˜LÐ(Ð(rC   rO   )r4   r5   r6   r7   r   rQ   re   rn   r>   r   r   rl   ro   rp   s   @rD   rr   rr   m   sÍ   ø€ € € € € ðð ðÐ/ð ð ð ð ð ð ð /3ð	,)ð ,)à”|ð,)ð # 5¤<°´Ð#=Ô>ð,)ð œ tÑ+ð	,)ð
 Ð+Ô,ð,)ð 
ˆuŒ|˜Uœ\¨DÑ0Ð0Ô	1ð,)ð ,)ð ,)ð ,)ð ,)ð ,)ð ,)ð ,)rC   rr   c                   ó¤   ‡ — e Zd Z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e	         deej
                 f
d	„Zˆ xZS )ÚMLCDEncoderLayerrM   c                 ór   •— t          ¦   «                              |¦  «         t          |¦  «        | _        d S rO   )rP   rQ   rr   Ú	self_attnrS   s     €rD   rQ   zMLCDEncoderLayer.__init__ª   s.   ø€ Ý‰Œ×Ò˜Ñ Ô Ð Ý& vÑ.Ô.ˆŒˆˆrC   Nru   rv   rw   rx   rW   c                 óÈ   — |}|                       |¦  «        } | j        d|||dœ|¤Ž\  }}||z   }|}|                      |¦  «        }|                      |¦  «        }||z   }|S )a¸  
        Args:
            hidden_states (`torch.FloatTensor`):
                Input to the layer of shape `(batch, seq_len, embed_dim)`.
                Represents the hidden states from the previous layer or the input embeddings.
            position_embeddings (`tuple[torch.Tensor, torch.Tensor]`):
                A tuple of two tensors, each of shape `(batch, seq_len, embed_dim)`.
                Represents absolute positional embeddings for the query and key in the attention mechanism.
            attention_mask (`torch.FloatTensor`):
                Attention mask of shape `(batch, 1, q_len, k_v_seq_len)` where padding elements are indicated by very large negative values.
        )ru   rv   rw   rB   )Úlayer_norm1r—   Úlayer_norm2Úmlp)rT   ru   rv   rw   rx   ÚresidualÚ_s          rD   rl   zMLCDEncoderLayer.forward®   s–   € ð$ !ˆà×(Ò(¨Ñ7Ô7ˆØ)˜4œ>ð 
Ø'Ø 3Ø)ð
ð 
ð ð	
ð 
Ñˆ�qð ! =Ñ0ˆà ˆØ×(Ò(¨Ñ7Ô7ˆØŸš Ñ/Ô/ˆØ  =Ñ0ˆàÐrC   rO   )r4   r5   r6   r   rQ   re   rn   r>   r   r   rm   rl   ro   rp   s   @rD   r•   r•   ©   s¸   ø€ € € € € ð/Ð/ð /ð /ð /ð /ð /ð /ð /3ð	"ð "à”|ð"ð # 5¤<°´Ð#=Ô>ð"ð œ tÑ+ð	"ð
 Ð+Ô,ð"ð 
ˆuÔ Ô	!ð"ð "ð "ð "ð "ð "ð "ð "rC   r•   c                   ó˜   ‡ — e Zd ZdZ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
e         d	eez  f
d
„Zˆ xZS )ÚMLCDEncoderz³
    Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a
    [`MLCDEncoderLayer`].

    Args:
        config: MLCDVisionConfig
    rM   c                 óJ   •— t          ¦   «                              |¦  «         dS )z3Overwrite dummy `MLCDConfig` to `MLCDVisionConfig`.N)rP   rQ   rS   s     €rD   rQ   zMLCDEncoder.__init__Ü   s!   ø€ å‰Œ×Ò˜Ñ Ô Ð Ð Ð rC   NÚinputs_embedsrv   rw   rx   rW   c                 óP   — |}| j         D ]} ||||fi |¤Ž}Œt          |¬¦  «        S )a=  
        Args:
            inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
                Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation.
                This is useful if you want more control over how to convert `input_ids` indices into associated vectors
                than the model's internal embedding lookup matrix.
            position_embeddings (`tuple[torch.Tensor, torch.Tensor]`):
                A tuple of two tensors, each of shape `(batch, seq_len, embed_dim)`.
                Represents absolute positional embeddings for the query and key in the attention mechanism.
            attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
                Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
                - 1 for tokens that are **not masked**,
                - 0 for tokens that are **masked**.
                [What are attention masks?](../glossary#attention-mask)
        )Úlast_hidden_state)Úlayersr   )rT   r¡   rv   rw   rx   ru   Úencoder_layers          rD   rl   zMLCDEncoder.forwardà   sa   € ð, &ˆØ!œ[ð 	ð 	ˆMØ)˜MØØ#Øðð ð ð	ð ˆMˆMõ Ø+ð
ñ 
ô 
ð 	
rC   rO   )r4   r5   r6   r7   r   rQ   re   rm   r>   rn   r   r   r   rl   ro   rp   s   @rD   rŸ   rŸ   Ó   sÀ   ø€ € € € € ðð ð!Ð/ð !ð !ð !ð !ð !ð !ð /3ð	!
ð !
àÔ(ð!
ð # 5¤<°´Ð#=Ô>ð!
ð œ tÑ+ð	!
ð
 Ð+Ô,ð!
ð 
�Ñ	 ð!
ð !
ð !
ð !
ð !
ð !
ð !
ð !
rC   rŸ   c                   ó€   ‡ — e Zd ZU eed<   dZdgZd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MLCDPreTrainedModelrM   Úvision_modelr•   TF)ru   Ú
attentionsc                 óŠ  •— t          ¦   «                              |¦  «         | j        j        }t	          |t
          ¦  «        r±| j        j        }t          j        |j        d|j	        dz  |z  ¬¦  «         t          j        |j
        j        |j        j        |z  ¬¦  «         t          j        |j        t          j        |j        j        d         ¦  «                             d¦  «        ¦  «         dS t	          |t&          ¦  «        r»| j        j        }|j	        dz  d|j        j        z  dz  z  |z  }|j	        dz  |z  }t          j        |j        j        |¬¦  «         t          j        |j        j        |¬¦  «         t          j        |j        j        |¬¦  «         t          j        |j        j        |¬¦  «         dS t	          |t2          ¦  «        rˆ| j        j        }|j        j        dz  d|j        j        z  dz  z  |z  }d|j        j        z  dz  |z  }t          j        |j        j        |¬¦  «         t          j        |j        j        |¬¦  «         dS t	          |t:          ¦  «        rL| j        j        }|j        j        |j        j        z  dz  dz  |z  }t          j        |j        d|¬¦  «         dS dS )	zInitialize the weightsr/   g      à¿)ÚmeanÚstd)r¬   rZ   )r&   rZ   r   N) rP   Ú_init_weightsrM   r2   Ú
isinstancerL   ÚinitÚnormal_rc   Ú	embed_dimr^   r_   r1   Úcopy_Úposition_idsre   Úaranger]   rd   rr   r#   r|   r€   r�   rŠ   rF   r    Úfc1Úfc2ÚMLCDVisionModelr%   Úclass_pos_emb)rT   ÚmoduleÚfactorÚin_proj_stdÚout_proj_stdÚfc_stdÚpos_emb_stdrU   s          €rD   r­   z!MLCDPreTrainedModel._init_weights  s¢  ø€ õ 	‰Œ×Ò˜fÑ%Ô%Ð%Ø”Ô/ˆÝ�fÕ2Ñ3Ô3ð 	JØ”[Ô3ˆFÝŒL˜Ô/°c¸vÔ?OÐQUÑ?UÐX^Ñ?^Ð_Ñ_Ô_Ð_ÝŒL˜Ô/Ô6¸F¼MÔ<[Ð^dÑ<dÐeÑeÔeÐeÝŒJ�vÔ*­E¬L¸Ô9LÔ9RÐSUÔ9VÑ,WÔ,W×,^Ò,^Ð_fÑ,gÔ,gÑhÔhÐhÐhÐhÝ˜¥Ñ.Ô.ð 	JØ”[Ô3ˆFØ!Ô+¨TÑ1°q¸6¼=Ô;ZÑ7ZÐ_cÑ6cÑdÐgmÑmˆKØ"Ô,¨dÑ2°fÑ<ˆLÝŒL˜œÔ-°;Ð?Ñ?Ô?Ð?ÝŒL˜œÔ-°;Ð?Ñ?Ô?Ð?ÝŒL˜œÔ-°;Ð?Ñ?Ô?Ð?ÝŒL˜œÔ/°\ÐBÑBÔBÐBÐBÐBÝ˜¥Ñ(Ô(ð 		JØ”[Ô3ˆFØ!œ=Ô4°dÑ:ÀÀFÄMÔDcÑ@cÐhlÑ?lÑmÐpvÑvˆKØ˜&œ-Ô3Ñ3¸Ñ<¸vÑEˆFÝŒL˜œÔ*°Ð7Ñ7Ô7Ð7ÝŒL˜œÔ*°Ð<Ñ<Ô<Ð<Ð<Ð<Ý˜¥Ñ0Ô0ð 	JØ”[Ô3ˆFØ!œ=Ô4¸¼Ô8YÑYÐ]^Ñ^ÐcgÑgÐjpÑpˆKÝŒL˜Ô-°C¸[ÐIÑIÔIÐIÐIÐIð	Jð 	JrC   )r4   r5   r6   r   r<   Úbase_model_prefixÚ_no_split_modulesÚsupports_gradient_checkpointingÚaccepts_loss_kwargsÚ_supports_flash_attnÚ_supports_sdpaÚ_supports_flex_attnÚ_supports_attention_backendr•   rr   Ú_can_record_outputsre   Úno_gradr­   ro   rp   s   @rD   r§   r§     s¨   ø€ € € € € € àÐÐÑØ&ÐØ+Ð,ÐØ&*Ð#ØÐØÐØ€NØÐØ"&Ðà)Ø#ðð Ðð
 €U„]�_„_ðJð Jð Jð Jñ „_ðJð Jð Jð Jð JrC   r§   c                   ó^   ‡ — e Zd Zdefˆ fd„Z	 ddej        dz  dee         de	e
z  fd„Zˆ xZS )	r·   rM   c                 óê   •— t          ¦   «                              |¦  «         t          |¦  «        | _        t	          j        t          j        d|j        |j	        z  dz  ¦  «        ¦  «        | _
        d S )Nr&   r   )rP   rQ   rJ   Úvision_rotary_embeddingÚnnÚ	Parameterre   Úrandnr    r%   r¸   rS   s     €rD   rQ   zMLCDVisionModel.__init__3  s`   ø€ Ý‰Œ×Ò˜Ñ Ô Ð Ý':¸6Ñ'BÔ'BˆÔ$Ýœ\­%¬+°a¸Ô9KÈvÔOiÑ9iÐmnÑ9nÑ*oÔ*oÑpÔpˆÔÐÐrC   NrV   rx   rW   c                 ó*  — |€t          d¦  «        ‚|                      |¦  «        }|                      |¦  «        }|j        d         | j        j        z  }|j        d         | j        j        z  }t          j        ||j        ¬¦  «         	                    d¦  «         
                    d|¦  «        }t          j        ||j        ¬¦  «         	                    d¦  «         
                    |d¦  «        }t          j        |                     ¦   «         |                     ¦   «         gd¬¦  «        }|                      ||¦  «        }	t          j        | j        | j        gd¬¦  «        }
t          j        |
                     ¦   «         |	d         gd¬¦  «        t          j        |
                     ¦   «         |	d         gd¬¦  «        g}	 | j        d||	d	œ|¤Ž}|d         }|dd…ddd…f         }|                      |¦  «        }t)          ||¬
¦  «        S )aû  
        Example:

        ```python
        >>> import httpx
        >>> from io import BytesIO
        >>> from PIL import Image
        >>> from transformers import AutoProcessor, MLCDVisionModel
        >>> model = MLCDVisionModel.from_pretrained("DeepGlint-AI/mlcd-vit-bigG-patch14-448")
        >>> processor = AutoProcessor.from_pretrained("DeepGlint-AI/mlcd-vit-bigG-patch14-448")

        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
        >>> with httpx.stream("GET", url) as response:
        ...     image = Image.open(BytesIO(response.read()))
        >>> inputs = processor(images=image, return_tensors="pt")

        >>> with torch.no_grad():
        ...     outputs = model(**inputs, output_attentions=True)

        >>> features = outputs.last_hidden_state
        >>> print(f"Extracted features shape: {features.shape}")
        >>> print(f"Number of attention layers: {len(outputs.attentions)}")
        >>> print(f"Attention shape: {outputs.attentions[0].shape}")
        ```Nz You have to specify pixel_valueséþÿÿÿrZ   )Údevicer&   r   r[   )r¡   rv   )r£   Úpooler_outputrB   )Ú
ValueErrorrk   Úpre_layrnormr]   rM   r+   re   r´   rÑ   r‚   rd   Ústackra   rË   rf   r¸   r�   r�   ÚencoderÚpost_layernormr	   )rT   rV   rx   ru   Únum_patches_heightÚnum_patches_widthÚhpos_idsÚwpos_idsÚpos_idsrv   r¸   Úencoder_outputsr£   Úpooled_outputs                 rD   rl   zMLCDVisionModel.forward8  s.  € ð: ÐÝÐ?Ñ@Ô@Ð@àŸš¨Ñ5Ô5ˆØ×)Ò)¨-Ñ8Ô8ˆà)Ô/°Ô3°t´{Ô7MÑMÐØ(Ô.¨rÔ2°d´kÔ6LÑLÐåŒLÐ+°LÔ4GÐHÑHÔH×RÒRÐSTÑUÔU×\Ò\Ð]_ÐarÑsÔsð 	õ ŒLÐ*°<Ô3FÐGÑGÔG×QÒQÐRSÑTÔT×[Ò[Ð\nÐprÑsÔsð 	õ ”+˜x×/Ò/Ñ1Ô1°8×3CÒ3CÑ3EÔ3EÐFÈBÐOÑOÔOˆØ"×:Ò:¸=È'ÑRÔRÐÝœ	 4Ô#5°tÔ7IÐ"JÐPRÐSÑSÔSˆåŒI�}×(Ò(Ñ*Ô*Ð,?ÀÔ,BÐCÈÐKÑKÔKÝŒI�}×(Ò(Ñ*Ô*Ð,?ÀÔ,BÐCÈÐKÑKÔKð
Ðð
 '˜$œ,ð 
Ø'Ø 3ð
ð 
ð ð
ð 
ˆð ,¨AÔ.ÐØ)¨!¨!¨!¨Q°°°¨'Ô2ˆØ×+Ò+¨MÑ:Ô:ˆå)Ø/Ø'ð
ñ 
ô 
ð 	
rC   rO   )r4   r5   r6   r   rQ   re   rm   r   r   r>   r	   rl   ro   rp   s   @rD   r·   r·   2  s¦   ø€ € € € € ðqÐ/ð qð qð qð qð qð qð 26ð@
ð @
àÔ'¨$Ñ.ð@
ð Ð+Ô,ð@
ð 
Ð+Ñ	+ð	@
ð @
ð @
ð @
ð @
ð @
ð @
ð @
rC   r·   )r   r§   r·   )1Úcollections.abcr   re   Útorch.nnrÌ   Úhuggingface_hub.dataclassesr   Ú r   r¯   Úconfiguration_utilsr   Úmodeling_outputsr   r	   Úmodeling_utilsr
   r   Úprocessing_utilsr   Úutilsr   r   r   Úclip.modeling_clipr   r   r   r   r   r   Úllama.modeling_llamar   Úqwen2_5_vl.modeling_qwen2_5_vlr   r   Ú
get_loggerr4   Úloggerr   rF   rJ   rL   rr   r•   rŸ   r§   r·   Ú__all__rB   rC   rD   ú<module>rî      s5  ðð %Ð $Ð $Ð $Ð $Ð $à €€€Ø Ð Ð Ð Ð Ð Ø .Ð .Ð .Ð .Ð .Ð .à &Ð &Ð &Ð &Ð &Ð &Ø 3Ð 3Ð 3Ð 3Ð 3Ð 3Ø KÐ KÐ KÐ KÐ KÐ KÐ KÐ KØ FÐ FÐ FÐ FÐ FÐ FÐ FÐ FØ &Ð &Ð &Ð &Ð &Ð &Ø @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @Ð @ðð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ;Ð :Ð :Ð :Ð :Ð :Ø iÐ iÐ iÐ iÐ iÐ iÐ iÐ ið 
ˆÔ	˜HÑ	%Ô	%€ð €ÐCÐDÑDÔDØð%(ð %(ð %(ð %(ð %(Ð'ñ %(ô %(ñ „ñ EÔDð%(ðP	ð 	ð 	ð 	ð 	ˆgñ 	ô 	ð 	ð	ð 	ð 	ð 	ð 	Ð9ñ 	ô 	ð 	ðð ð ð ð Ð/ñ ô ð ð$9)ð 9)ð 9)ð 9)ð 9)�Mñ 9)ô 9)ð 9)ðx'ð 'ð 'ð 'ð 'Ð'ñ 'ô 'ð 'ðT.
ð .
ð .
ð .
ð .
�+ñ .
ô .
ð .
ðb ð*Jð *Jð *Jð *Jð *J˜/ñ *Jô *Jñ „ð*JðZF
ð F
ð F
ð F
ð F
�oñ F
ô F
ð F
ðRð ð €€€rC   