§
    ·¯¸jÕ]  ã            	       óº  — d Z ddlZddlmZmZmZmZmZmZm	Z	m
Z
 ddlZddlmZ ddlmc mZ ddl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 dd	lm Z  dd
l!m"Z"m#Z# dgZ$ G d„ dej%        ¦  «        Z& G d„ dej%        ¦  «        Z' G d„ dej%        ¦  «        Z( G d„ dej%        ¦  «        Z) G d„ dej%        ¦  «        Z* G d„ dej%        ¦  «        Z+ G d„ dej%        ¦  «        Z, G d„ dej%        ¦  «        Z- G d„ dej%        ¦  «        Z.dee/ej0        f         dej%        dee/ej0        f         fd „Z1d0d"e/d#edee/ef         fd$„Z2 e" e2d%¬&¦  «         e2d%¬&¦  «         e2d%¬&¦  «         e2d%¬&¦  «        d'œ¦  «        Z3d1d)e/d*e4d#ede.fd+„Z5e#d1d*e4d#ede.fd,„¦   «         Z6e#d1d*e4d#ede.fd-„¦   «         Z7e#d1d*e4d#ede.fd.„¦   «         Z8e#d1d*e4d#ede.fd/„¦   «         Z9dS )2aj  SwiftFormer
SwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision Applications
Code: https://github.com/Amshaker/SwiftFormer
Paper: https://arxiv.org/pdf/2303.15446

@InProceedings{Shaker_2023_ICCV,
    author    = {Shaker, Abdelrahman and Maaz, Muhammad and Rasheed, Hanoona and Khan, Salman and Yang, Ming-Hsuan and Khan, Fahad Shahbaz},
    title     = {SwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision Applications},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    year      = {2023},
}
é    N)ÚAnyÚDictÚListÚOptionalÚSetÚTupleÚTypeÚUnion©ÚIMAGENET_DEFAULT_MEANÚIMAGENET_DEFAULT_STD)ÚDropPathÚLinearÚ	LayerTypeÚ	to_2tupleÚtrunc_normal_Úget_device_dtypeé   )Úbuild_model_with_cfg)Úfeature_take_indices)Úcheckpoint_seq)Úgenerate_default_cfgsÚregister_modelÚSwiftFormerc                   óR   ‡ — e Zd Zddededefˆ fd„Zdej        d	ej        fd
„Z	ˆ xZ
S )ÚLayerScale2dçñhãˆµøä>FNÚdimÚinit_valuesÚinplacec                 óÀ   •— ||dœ}t          ¦   «                              ¦   «          || _        t          j        |t          j        |ddfi |¤Žz  d¬¦  «        | _        d S )N©ÚdeviceÚdtyper   T)Úrequires_grad)ÚsuperÚ__init__r    ÚnnÚ	ParameterÚtorchÚonesÚgamma)Úselfr   r   r    r#   r$   ÚddÚ	__class__s          €úg/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/timm/models/swiftformer.pyr'   zLayerScale2d.__init__   sl   ø€ Ø¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØˆŒÝ”\Ø�%œ* S¨!¨QÐ5Ð5°"Ð5Ð5Ñ5ÀTðKñ Kô KˆŒ
ˆ
ˆ
ó    ÚxÚreturnc                 óX   — | j         r|                     | j        ¦  «        n	|| j        z  S ©N)r    Úmul_r,   ©r-   r2   s     r0   ÚforwardzLayerScale2d.forward&   s(   € Ø%)¤\ÐEˆq�vŠv�d”jÑ!Ô!Ð!°q¸4¼:±~ÐEr1   )r   FNN)Ú__name__Ú
__module__Ú__qualname__ÚintÚfloatÚboolr'   r*   ÚTensorr8   Ú__classcell__©r/   s   @r0   r   r      s‘   ø€ € € € € ðKð K˜Cð K¨eð KÀTð Kð Kð Kð Kð Kð KðF˜œð F¨%¬,ð Fð Fð Fð Fð Fð Fð Fð Fr1   r   c                   ó’   ‡ — e Zd ZdZdddddej        ddfdeded	ed
ededeej                 fˆ fd„Z	de
j        de
j        fd„Zˆ xZS )Ú	Embeddingz 
    Patch Embedding that is implemented by a layer of conv.
    Input: tensor in shape [B, C, H, W]
    Output: tensor in shape [B, C, H/stride, W/stride]
    é   i   é   r   NÚin_chansÚ	embed_dimÚ
patch_sizeÚstrideÚpaddingÚ
norm_layerc	                 ó(  •— ||dœ}	t          ¦   «                              ¦   «          t          |¦  «        }t          |¦  «        }t          |¦  «        }t          j        |||||fi |	¤Ž| _        |r	 ||fi |	¤Žnt          j        ¦   «         | _        d S )Nr"   )r&   r'   r   r(   ÚConv2dÚprojÚIdentityÚnorm)r-   rF   rG   rH   rI   rJ   rK   r#   r$   r.   r/   s             €r0   r'   zEmbedding.__init__0   sž   ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐÝ˜zÑ*Ô*ˆ
Ý˜6Ñ"Ô"ˆÝ˜GÑ$Ô$ˆÝ”I˜h¨	°:¸vÀwÐUÐUÐRTÐUÐUˆŒ	Ø3=ÐP�J�J˜yÐ/Ð/¨BÐ/Ð/Ð/Å2Ä;Á=Ä=ˆŒ	ˆ	ˆ	r1   r2   r3   c                 óZ   — |                       |¦  «        }|                      |¦  «        }|S r5   )rN   rP   r7   s     r0   r8   zEmbedding.forwardC   s%   € Ø�IŠI�a‰LŒLˆØ�IŠI�a‰LŒLˆØˆr1   )r9   r:   r;   Ú__doc__r(   ÚBatchNorm2dr<   r	   ÚModuler'   r*   r?   r8   r@   rA   s   @r0   rC   rC   *   sÛ   ø€ € € € € ðð ð Ø Ø ØØØ*,¬.ØØðQð QàðQð ðQð ð	Qð
 ðQð ðQð ˜RœYœðQð Qð Qð Qð Qð Qð&˜œð ¨%¬,ð ð ð ð ð ð ð ð r1   rC   c                   ó¶   ‡ — e Zd ZdZdddej        ej        dddfdeded	ed
ede	ej
                 de	ej
                 defˆ fd„Zdej        dej        fd„Zˆ xZS )ÚConvEncoderz›
    Implementation of ConvEncoder with 3*3 and 1*1 convolutions.
    Input: tensor with shape [B, C, H, W]
    Output: tensor with shape [B, C, H, W]
    é@   rD   ç        TNr   Ú
hidden_dimÚkernel_sizeÚ	drop_pathÚ	act_layerrK   Úuse_layer_scalec
                 óØ  •— ||	dœ}
t          ¦   «                              ¦   «          t          j        |||f|dz  |dœ|
¤Ž| _         ||fi |
¤Ž| _        t          j        ||dfi |
¤Ž| _         |¦   «         | _        t          j        ||dfi |
¤Ž| _        |dk    rt          |¦  «        nt          j
        ¦   «         | _        |rt          |dfi |
¤Žnt          j
        ¦   «         | _        d S )Nr"   é   ©rJ   Úgroupsr   rX   ©r&   r'   r(   rM   ÚdwconvrP   Úpwconv1ÚactÚpwconv2r   rO   r[   r   Úlayer_scale)r-   r   rY   rZ   r[   r\   rK   r]   r#   r$   r.   r/   s              €r0   r'   zConvEncoder.__init__O   s  ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐÝ”i  S¨+Ðb¸{ÈaÑ?OÐX[ÐbÐbÐ_aÐbÐbˆŒØ�J˜sÐ)Ð) bÐ)Ð)ˆŒ	Ý”y  j°!Ð:Ð:°rÐ:Ð:ˆŒØ�9‘;”;ˆŒÝ”y ¨S°!Ð:Ð:°rÐ:Ð:ˆŒØ09¸B²°� )Ñ,Ô,Ð,ÅBÄKÁMÄMˆŒØ9HÐ[�<¨¨QÐ5Ð5°"Ð5Ð5Ð5ÍbÌkÉmÌmˆÔÐÐr1   r2   r3   c                 ó6  — |}|                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }||                      |¦  «        z   }|S r5   ©rc   rP   rd   re   rf   rg   r[   )r-   r2   Úinputs      r0   r8   zConvEncoder.forwarde   s€   € ØˆØ�KŠK˜‰NŒNˆØ�IŠI�a‰LŒLˆØ�LŠL˜‰OŒOˆØ�HŠH�Q‰KŒKˆØ�LŠL˜‰OŒOˆØ×Ò˜QÑÔˆØ�D—N’N 1Ñ%Ô%Ñ%ˆØˆr1   ©r9   r:   r;   rR   r(   ÚGELUrS   r<   r=   r	   rT   r>   r'   r*   r?   r8   r@   rA   s   @r0   rV   rV   I   sî   ø€ € € € € ðð ð !Ø Ø!Ø)+¬Ø*,¬.Ø$(ØØð\ð \àð\ð ð\ð ð	\ð
 ð\ð ˜BœI”ð\ð ˜RœYœð\ð "ð\ð \ð \ð \ð \ð \ð,	˜œð 	¨%¬,ð 	ð 	ð 	ð 	ð 	ð 	ð 	ð 	r1   rV   c                   óÈ   ‡ — e Zd ZdZddej        ej        dddfdedee         dee         de	ej
                 de	ej
                 d	efˆ fd
„Zdej        dej        fd„Zˆ xZS )ÚMlpz‘
    Implementation of MLP layer with 1*1 convolutions.
    Input: tensor with shape [B, C, H, W]
    Output: tensor with shape [B, C, H, W]
    NrX   Úin_featuresÚhidden_featuresÚout_featuresr\   rK   Údropc	                 ó2  •— ||dœ}	t          ¦   «                              ¦   «          |p|}|p|} ||fi |	¤Ž| _        t          j        ||dfi |	¤Ž| _         |¦   «         | _        t          j        ||dfi |	¤Ž| _        t          j        |¦  «        | _	        d S )Nr"   r   )
r&   r'   Únorm1r(   rM   Úfc1re   Úfc2ÚDropoutrr   )r-   ro   rp   rq   r\   rK   rr   r#   r$   r.   r/   s             €r0   r'   zMlp.__init__w   s±   ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ#Ð2 {ˆØ)Ð8¨[ˆØ�Z Ð2Ð2¨rÐ2Ð2ˆŒ
Ý”9˜[¨/¸1ÐCÐCÀÐCÐCˆŒØ�9‘;”;ˆŒÝ”9˜_¨l¸AÐDÐDÀÐDÐDˆŒÝ”J˜tÑ$Ô$ˆŒ	ˆ	ˆ	r1   r2   r3   c                 ó  — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|S r5   )rt   ru   re   rr   rv   r7   s     r0   r8   zMlp.forwardŒ   sa   € Ø�JŠJ�q‰MŒMˆØ�HŠH�Q‰KŒKˆØ�HŠH�Q‰KŒKˆØ�IŠI�a‰LŒLˆØ�HŠH�Q‰KŒKˆØ�IŠI�a‰LŒLˆØˆr1   )r9   r:   r;   rR   r(   rl   rS   r<   r   r	   rT   r=   r'   r*   r?   r8   r@   rA   s   @r0   rn   rn   q   sÛ   ø€ € € € € ðð ð .2Ø*.Ø)+¬Ø*,¬.ØØØð%ð %àð%ð & cœ]ð%ð # 3œ-ð	%ð
 ˜BœI”ð%ð ˜RœYœð%ð ð%ð %ð %ð %ð %ð %ð*˜œð ¨%¬,ð ð ð ð ð ð ð ð r1   rn   c                   ó`   ‡ — e Zd ZdZ	 	 	 	 	 ddededefˆ fd	„Zd
ej        dej        fd„Zˆ xZ	S )ÚEfficientAdditiveAttentionz�
    Efficient Additive Attention module for SwiftFormer.
    Input: tensor in shape [B, C, H, W]
    Output: tensor in shape [B, C, H, W]
    é   é   r   NÚin_dimsÚ	token_dimÚ	num_headsc                 óž  •— ||dœ}t          ¦   «                              ¦   «          |dz  | _        t          j        |||z  fi |¤Ž| _        t          j        |||z  fi |¤Ž| _        t          j        t          j	        ||z  dfi |¤Ž¦  «        | _
        t          j        ||z  ||z  fi |¤Ž| _        t          j        ||z  |fi |¤Ž| _        d S )Nr"   g      à¿r   )r&   r'   Úscale_factorr(   r   Úto_queryÚto_keyr)   r*   ÚrandnÚw_grN   Úfinal)r-   r}   r~   r   r#   r$   r.   r/   s          €r0   r'   z#EfficientAdditiveAttention.__init__œ   sè   ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ%¨Ñ-ˆÔÝœ	 '¨9°yÑ+@ÐGÐGÀBÐGÐGˆŒÝ”i ¨°YÑ)>ÐEÐEÀ"ÐEÐEˆŒå”<¥¤¨I¸	Ñ,AÀ1Ð KÐ KÈÐ KÐ KÑLÔLˆŒå”I˜i¨)Ñ3°YÀÑ5JÐQÐQÈbÐQÐQˆŒ	Ý”Y˜y¨9Ñ4°iÐFÐFÀ2ÐFÐFˆŒ
ˆ
ˆ
r1   r2   r3   c                 óL  — |j         \  }}}}|                     d¦  «                             ddd¦  «        }t          j        |                      |¦  «        d¬¦  «        }t          j        |                      |¦  «        d¬¦  «        }t          j        || j        z  | j        z  d¬¦  «        }t          j
        ||z  dd¬¦  «        }|                      ||z  ¦  «        |z   }	|                      |	¦  «                             ddd¦  «                             |d||¦  «        }	|	S )Nr_   r   r   éÿÿÿÿ©r   T)r   Úkeepdim)ÚshapeÚflattenÚpermuteÚFÚ	normalizer‚   rƒ   r…   r�   r*   ÚsumrN   r†   Úreshape)
r-   r2   ÚBÚ_ÚHÚWÚqueryÚkeyÚattnÚouts
             r0   r8   z"EfficientAdditiveAttention.forward¯   s  € Ø”W‰
ˆˆ1ˆa�Ø�IŠI�a‰LŒL× Ò   A qÑ)Ô)ˆå”˜DŸMšM¨!Ñ,Ô,°"Ð5Ñ5Ô5ˆÝŒk˜$Ÿ+š+ a™.œ.¨bÐ1Ñ1Ô1ˆåŒ{˜5 4¤8Ñ+¨dÔ.?Ñ?ÀQÐGÑGÔGˆÝŒy˜ ™¨1°dÐ;Ñ;Ô;ˆà�iŠi˜˜s™
Ñ#Ô# eÑ+ˆØ�jŠj˜‰oŒo×%Ò% a¨¨AÑ.Ô.×6Ò6°q¸"¸aÀÑCÔCˆØˆ
r1   )r{   r|   r   NN)
r9   r:   r;   rR   r<   r'   r*   r?   r8   r@   rA   s   @r0   rz   rz   –   s©   ø€ € € € € ðð ð Ø ØØØðGð GàðGð ðGð ð	Gð Gð Gð Gð Gð Gð&˜œð ¨%¬,ð ð ð ð ð ð ð ð r1   rz   c                   ó°   ‡ — e Zd ZdZdddej        ej        ddfdededed	e	d
e
ej                 de
ej                 fˆ fd„Zdej        dej        fd„Zˆ xZS )ÚLocalRepresentationzÉ
    Local Representation module for SwiftFormer that is implemented by 3*3 depth-wise and point-wise convolutions.
    Input: tensor in shape [B, C, H, W]
    Output: tensor in shape [B, C, H, W]
    rD   rX   TNr   rZ   r[   r]   r\   rK   c	                 óÜ  •— ||dœ}	t          ¦   «                              ¦   «          t          j        |||f|dz  |dœ|	¤Ž| _         ||fi |	¤Ž| _        t          j        ||fddi|	¤Ž| _         |¦   «         | _        t          j        ||fddi|	¤Ž| _        |dk    rt          |¦  «        nt          j
        ¦   «         | _        |rt          |dfi |	¤Žnt          j
        ¦   «         | _        d S )Nr"   r_   r`   rZ   r   rX   rb   )r-   r   rZ   r[   r]   r\   rK   r#   r$   r.   r/   s             €r0   r'   zLocalRepresentation.__init__Ä   s  ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐÝ”i  S¨+Ðb¸{ÈaÑ?OÐX[ÐbÐbÐ_aÐbÐbˆŒØ�J˜sÐ)Ð) bÐ)Ð)ˆŒ	Ý”y  cÐ?Ð?°qÐ?¸BÐ?Ð?ˆŒØ�9‘;”;ˆŒÝ”y  cÐ?Ð?°qÐ?¸BÐ?Ð?ˆŒØ09¸B²°� )Ñ,Ô,Ð,ÅBÄKÁMÄMˆŒØ9HÐ[�<¨¨QÐ5Ð5°"Ð5Ð5Ð5ÍbÌkÉmÌmˆÔÐÐr1   r2   r3   c                 ó6  — |}|                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }||                      |¦  «        z   }|S r5   ri   )r-   r2   Úskips      r0   r8   zLocalRepresentation.forwardÙ   s€   € ØˆØ�KŠK˜‰NŒNˆØ�IŠI�a‰LŒLˆØ�LŠL˜‰OŒOˆØ�HŠH�Q‰KŒKˆØ�LŠL˜‰OŒOˆØ×Ò˜QÑÔˆØ�4—>’> !Ñ$Ô$Ñ$ˆØˆr1   )r9   r:   r;   rR   r(   rl   rS   r<   r=   r>   r	   rT   r'   r*   r?   r8   r@   rA   s   @r0   r›   r›   ¾   sà   ø€ € € € € ðð ð  !Ø!Ø$(Ø)+¬Ø*,¬.ØØð\ð \àð\ð ð\ð ð	\ð
 "ð\ð ˜BœI”ð\ð ˜RœYœð\ð \ð \ð \ð \ð \ð*	˜œð 	¨%¬,ð 	ð 	ð 	ð 	ð 	ð 	ð 	ð 	r1   r›   c                   ó¼   ‡ — e Zd ZdZdddej        ej        ddddf	deded	ed
ede	ej
                 de	ej
                 dedefˆ fd„Zdej        dej        fd„Zˆ xZS )ÚBlockzî
    SwiftFormer Encoder Block for SwiftFormer. It consists of :
    (1) Local representation module, (2) EfficientAdditiveAttention, and (3) MLP block.
    Input: tensor in shape [B, C, H, W]
    Output: tensor in shape [B, C, H, W]
    ç      @rX   Tr   Nr   Ú	mlp_ratioÚ	drop_rater[   r\   rK   r]   Úlayer_scale_init_valuec           	      óî  •— |	|
dœ}t          ¦   «                              ¦   «          t          d||||dœ|¤Ž| _        t	          d||dœ|¤Ž| _        t          d|t          ||z  ¦  «        |||dœ|¤Ž| _        |dk    rt          |¦  «        nt          j        ¦   «         | _        |rt          ||fi |¤Žnt          j        ¦   «         | _        |rt          ||fi |¤Žnt          j        ¦   «         | _        d S )Nr"   )r   r]   r\   rK   )r}   r~   )ro   rp   r\   rK   rr   rX   © )r&   r'   r›   Úlocal_representationrz   r˜   rn   r<   Úlinearr   r(   rO   r[   r   Úlayer_scale_1Úlayer_scale_2)r-   r   r¢   r£   r[   r\   rK   r]   r¤   r#   r$   r.   r/   s               €r0   r'   zBlock.__init__ì   sF  ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐÝ$7ð %
ØØ+ØØ!ð	%
ð %
ð
 ð%
ð %
ˆÔ!õ /ÐP°sÀcÐPÐPÈRÐPÐPˆŒ	Ýð 
ØÝ  i¡Ñ0Ô0ØØ!Øð
ð 
ð ð
ð 
ˆŒð 1:¸B²°� )Ñ,Ô,Ð,ÅBÄKÁMÄMˆŒàð2�\¨#Ð/EÐLÐLÈÐLÐLÐLÝ$&¤K¡M¤Mð 	Ôð ð2�\¨#Ð/EÐLÐLÈÐLÐLÐLÝ$&¤K¡M¤Mð 	ÔÐÐr1   r2   r3   c                 ó(  — |                       |¦  «        }||                      |                      |                      |¦  «        ¦  «        ¦  «        z   }||                      |                      |                      |¦  «        ¦  «        ¦  «        z   }|S r5   )r§   r[   r©   r˜   rª   r¨   r7   s     r0   r8   zBlock.forward  sv   € Ø×%Ò% aÑ(Ô(ˆØ�—’˜t×1Ò1°$·)²)¸A±,´,Ñ?Ô?Ñ@Ô@Ñ@ˆØ�—’˜t×1Ò1°$·+²+¸a±.´.ÑAÔAÑBÔBÑBˆØˆr1   rk   rA   s   @r0   r    r    å   sî   ø€ € € € € ðð ð  "Ø!Ø!Ø)+¬Ø*,¬.Ø$(Ø,0ØØð#2ð #2àð#2ð ð#2ð ð	#2ð
 ð#2ð ˜BœI”ð#2ð ˜RœYœð#2ð "ð#2ð %*ð#2ð #2ð #2ð #2ð #2ð #2ðJ˜œð ¨%¬,ð ð ð ð ð ð ð ð r1   r    c                   óø   ‡ — e Zd ZdZdej        ej        dddddddf
deded	ee         d
e	de
ej                 de
ej                 de	de	dede	dee
ej                          fˆ fd„Zdej        dej        fd„Zˆ xZS )ÚStagezö
    Implementation of each SwiftFormer stages. Here, SwiftFormerEncoder used as the last block in all stages, while ConvEncoder used in the rest of the blocks.
    Input: tensor in shape [B, C, H, W]
    Output: tensor in shape [B, C, H, W]
    r¡   rX   Tr   Nr   ÚindexÚlayersr¢   r\   rK   r£   Údrop_path_rater]   r¤   Ú
downsamplec                 ó0  •— ||dœ}t          ¦   «                              ¦   «          d| _        |�|nt          j        ¦   «         | _        g }t          ||         ¦  «        D ]¡}||t          |d |…         ¦  «        z   z  t          |¦  «        dz
  z  }||         |z
  dk    r)|                     t          |f||||||	|
dœ|¤Ž¦  «         Œj|                     t          d|t          ||z  ¦  «        d||||	dœ|¤Ž¦  «         Œ¢t          j        |Ž | _        d S )Nr"   Fr   )r¢   r£   r[   r\   rK   r]   r¤   rD   )r   rY   rZ   r[   r\   rK   r]   r¦   )r&   r'   Úgrad_checkpointingr(   rO   r±   Úranger�   Úappendr    rV   r<   Ú
SequentialÚblocks)r-   r   r®   r¯   r¢   r\   rK   r£   r°   r]   r¤   r±   r#   r$   r.   r·   Ú	block_idxÚ	block_dprr/   s                     €r0   r'   zStage.__init__  so  ø€ ð  ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ"'ˆÔØ(2Ð(>˜*˜*ÅBÄKÁMÄMˆŒàˆÝ˜v eœ}Ñ-Ô-ð 	ð 	ˆIØ&¨)µc¸&ÀÀ%À¼.Ñ6IÔ6IÑ*IÑJÍcÐRXÉkÌkÐ\]ÉoÑ^ˆIØ�eŒ}˜yÑ(¨AÒ-Ð-Ø—’�eØð
à'Ø'Ø'Ø'Ø)Ø$3Ø+Að
ð 
ð ð
ð 
ñ 
ô 
ð 
ð 
ð —’�kð 	ØÝ" 9¨s¡?Ñ3Ô3Ø !Ø'Ø'Ø)Ø$3ð	ð 	ð ð	ð 	ñ 	ô 	ð 	ð 	õ ”m VÐ,ˆŒˆˆr1   r2   r3   c                 óÐ   — |                       |¦  «        }| j        r4t          j                             ¦   «         st          | j        |¦  «        }n|                      |¦  «        }|S r5   )r±   r³   r*   ÚjitÚis_scriptingr   r·   r7   s     r0   r8   zStage.forwardO  sZ   € Ø�OŠO˜AÑÔˆØÔ"ð 	­5¬9×+AÒ+AÑ+CÔ+Cð 	Ý˜tœ{¨AÑ.Ô.ˆAˆAà—’˜A‘”ˆAØˆr1   )r9   r:   r;   rR   r(   rl   rS   r<   r   r=   r	   rT   r>   r   r'   r*   r?   r8   r@   rA   s   @r0   r­   r­     s  ø€ € € € € ðð ð  "Ø)+¬Ø*,¬.Ø!Ø$&Ø$(Ø,0Ø48ØØð/-ð /-àð/-ð ð/-ð ˜”Ið	/-ð
 ð/-ð ˜BœI”ð/-ð ˜RœYœð/-ð ð/-ð "ð/-ð "ð/-ð %*ð/-ð !  b¤i¤Ô1ð/-ð /-ð /-ð /-ð /-ð /-ðb˜œð ¨%¬,ð ð ð ð ð ð ð ð r1   r­   c            !       ón  ‡ — e Zd Zg d¢g d¢dg d¢ej        ddddd	d	d
ddddddfdee         dee         dedee         deej	                 dedededede
de
dede
dededef ˆ fd„Zd „ Zej        j        d!efd"„¦   «         Zej        j        d:d$ed!eeef         fd%„¦   «         Zej        j        d;d&efd'„¦   «         Zej        j        d!eej	        ej	        f         fd(„¦   «         Zd<dedee         fd)„Zej        j        d;d&efd*„¦   «         Z	 	 	 	 	 d=d,ej        d-eeeee         f                  d.ed/ed0ed1ed!eeej                 eej        eej                 f         f         fd2„Z	 	 	 d>d-eeee         f         d3ed4efd5„Zd,ej        d!ej        fd6„Z d:d,ej        d7efd8„Z!d,ej        fd9„Z"ˆ xZ#S )?r   ©rD   rD   é   é   ©é0   é8   ép   éÜ   rÀ   )FTTTrD   r_   r   éè  rX   Tr   Úavgé    Nr¯   Ú
embed_dimsÚ
mlp_ratiosÚdownsamplesr\   Údown_patch_sizeÚdown_strideÚdown_padÚnum_classesr£   r°   r]   r¤   Úglobal_poolÚoutput_striderF   c                 ó  •— t          ¦   «                              ¦   «          ||dœ}|dk    sJ ‚|	| _        || _        || _        g | _        t          j        t          j        ||d         dz  dddfi |¤Žt          j	        |d         dz  fi |¤Žt          j
        ¦   «         t          j        |d         dz  |d         dddfi |¤Žt          j	        |d         fi |¤Žt          j
        ¦   «         ¦  «        | _        |d         }g }t          t          |¦  «        ¦  «        D ]Ÿ}||         rt          d|||         |||dœ|¤Žnt          j        ¦   «         }t!          d||         |||||
||||dœ
|¤Ž}||         }|                     |¦  «         | xj        t%          ||         d|dz   z  d	|› �¬
¦  «        gz  c_        Œ t          j        |Ž | _        |d         x| _        x| _        }t          j	        |fi |¤Ž| _        t          j        |
¦  «        | _        |	dk    rt3          ||	fi |¤Žnt          j        ¦   «         | _        |	dk    rt3          ||	fi |¤Žnt          j        ¦   «         | _        d| _        |                      ¦   «          d S )Nr"   rÈ   r   r_   rD   r   )rF   rG   rH   rI   rJ   )
r   r®   r¯   r¢   r\   r£   r°   r]   r¤   r±   ústages.)Únum_chsÚ	reductionÚmodulerˆ   Fr¦   )r&   r'   rÏ   rF   rÐ   Úfeature_infor(   r¶   rM   rS   ÚReLUÚstemr´   ÚlenrC   rO   r­   rµ   ÚdictÚstagesÚnum_featuresÚhead_hidden_sizerP   rw   Ú	head_dropr   ÚheadÚ	head_distÚdistilled_trainingÚ_initialize_weights)r-   r¯   rÉ   rÊ   rË   r\   rÌ   rÍ   rÎ   rÏ   r£   r°   r]   r¤   rÐ   rÑ   rF   r#   r$   Úkwargsr.   Úprev_dimrÜ   Úir±   ÚstageÚout_chsr/   s                              €r0   r'   zSwiftFormer.__init__Y  sý  ø€ õ, 	‰Œ×ÒÑÔÐØ¨Ð/Ð/ˆØ Ò"Ð"Ð"Ð"Ø&ˆÔØ ˆŒØ&ˆÔØˆÔå”MÝŒI�h 
¨1¤°Ñ 2°A°q¸!ÐBÐB¸rÐBÐBÝŒN˜: aœ=¨AÑ-Ð4Ð4°Ð4Ð4ÝŒG‰IŒIÝŒI�j ”m qÑ(¨*°Q¬-¸¸A¸qÐGÐGÀBÐGÐGÝŒN˜: aœ=Ð/Ð/¨BÐ/Ð/ÝŒG‰IŒIñ
ô 
ˆŒ	ð ˜a”=ˆàˆÝ•s˜6‘{”{Ñ#Ô#ð 	ið 	iˆAð ˜Q”ð3�ð Ø!Ø$ Qœ-Ø*Ø"Ø ðð ð ðð ð õ &(¤[¡]¤]ð õ ð Ø˜q”MØØØ$Ø#Ø#Ø-Ø /Ø'=Ø%ðð ð ðð ˆEð " !”}ˆHØ�MŠM˜%Ñ Ô Ð ØÐÔ¥$¨z¸!¬}ÈÈAÈaÉCÉÐYfÐcdÐYfÐYfÐ"gÑ"gÔ"gÐ!hÑhÐÔÐÝ”m VÐ,ˆŒð @JÈ"¼~ÐMˆÔÐM˜TÔ2°WÝ”N 7Ð1Ð1¨bÐ1Ð1ˆŒ	Ýœ IÑ.Ô.ˆŒØ:EÈº/¸/•F˜7 KÐ6Ð6°2Ð6Ð6Ð6ÍrÌ{É}Ì}ˆŒ	à?JÈQº¸� ¨Ð;Ð;¸Ð;Ð;Ð;ÕTVÔT_ÑTaÔTaˆŒØ"'ˆÔØ× Ò Ñ"Ô"Ð"Ð"Ð"r1   c                 ó¬  — |                       ¦   «         D ]¾\  }}t          |t          j        ¦  «        rCt	          |j        d¬¦  «         |j        �%t          j                             |j        d¦  «         Œbt          |t          j	        ¦  «        rBt	          |j        d¬¦  «         |j        �%t          j                             |j        d¦  «         Œ¿d S )Ng{®Gáz”?)Ústdr   )
Únamed_modulesÚ
isinstancer(   r   r   ÚweightÚbiasÚinitÚ	constant_rM   )r-   ÚnameÚms      r0   rã   zSwiftFormer._initialize_weights§  sÃ   € Ø×)Ò)Ñ+Ô+ð 	1ð 	1‰GˆD�!Ý˜!�RœYÑ'Ô'ð 1Ý˜aœh¨CÐ0Ñ0Ô0Ð0Ø”6Ð%Ý”G×%Ò% a¤f¨aÑ0Ô0Ð0øÝ˜A�rœyÑ)Ô)ð 1Ý˜aœh¨CÐ0Ñ0Ô0Ð0Ø”6Ð%Ý”G×%Ò% a¤f¨aÑ0Ô0Ð0øð	1ð 	1r1   r3   c                 ó   — t          ¦   «         S r5   )Úset©r-   s    r0   Úno_weight_decayzSwiftFormer.no_weight_decay²  s   € å‰uŒuˆr1   FÚcoarsec                 ó4   — t          d|rdng d¢¬¦  «        }|S )Nz^stemz^stages\.(\d+)))z^stages\.(\d+).downsample)r   )z^stages\.(\d+)\.blocks\.(\d+)N)z^norm)iŸ† )rÙ   r·   )rÛ   )r-   r÷   Úmatchers      r0   Úgroup_matcherzSwiftFormer.group_matcher¶  s=   € åØØ(.ð Ð$Ð$ð 5ð 5ð 5ð
ñ 
ô 
ˆð ˆr1   Úenablec                 ó(   — | j         D ]	}||_        Œ
d S r5   )rÜ   r³   )r-   rû   Úss      r0   Úset_grad_checkpointingz"SwiftFormer.set_grad_checkpointingÂ  s(   € à”ð 	*ð 	*ˆAØ#)ˆAÔ Ð ð	*ð 	*r1   c                 ó   — | j         | j        fS r5   ©rà   rá   rõ   s    r0   Úget_classifierzSwiftFormer.get_classifierÇ  s   € àŒy˜$œ.Ð(Ð(r1   c                 ó„  — t          | ¦  «        }|| _        |�|| _        |dk    rt          | j        |fi |¤Žnt          j        ¦   «         | _        | j                             | j	        ¦  «         |dk    rt          | j        |fi |¤Žnt          j        ¦   «         | _
        | j
                             | j	        ¦  «         d S )Nr   )r   rÏ   rÐ   r   rÝ   r(   rO   rà   ÚtrainÚtrainingrá   )r-   rÏ   rÐ   r.   s       r0   Úreset_classifierzSwiftFormer.reset_classifierË  sÁ   € Ý˜dÑ#Ô#ˆØ&ˆÔØÐ"Ø*ˆDÔØDOÐRSÂOÀO•F˜4Ô,¨kÐ@Ð@¸RÐ@Ð@Ð@ÕY[ÔYdÑYfÔYfˆŒ	ØŒ	�Š˜œÑ&Ô&Ð&ØITÐWXÊÈ� Ô 1°;ÐEÐEÀ"ÐEÐEÐEÕ^`Ô^iÑ^kÔ^kˆŒØŒ×Ò˜Tœ]Ñ+Ô+Ð+Ð+Ð+r1   c                 ó   — || _         d S r5   )râ   )r-   rû   s     r0   Úset_distilled_trainingz"SwiftFormer.set_distilled_trainingÕ  s   € à"(ˆÔÐÐr1   ÚNCHWr2   ÚindicesrP   Ú
stop_earlyÚ
output_fmtÚintermediates_onlyc                 ó0  — |dv s
J d¦   «         ‚g }t          t          | j        ¦  «        |¦  «        \  }}	t          | j        ¦  «        dz
  }
|                      |¦  «        }t          j                             ¦   «         s|s| j        }n| j        d|	dz   …         }t          |¦  «        D ]I\  }} ||¦  «        }||v r5|r||
k    r|                      |¦  «        }n|}| 	                    |¦  «         ŒJ|r|S ||
k    r|                      |¦  «        }||fS )aï   Forward features that returns intermediates.

        Args:
            x: Input image tensor
            indices: Take last n blocks if int, all if None, select matching indices if sequence
            norm: Apply norm layer to compatible intermediates
            stop_early: Stop iterating over blocks when last desired intermediate hit
            output_fmt: Shape of intermediate feature outputs
            intermediates_only: Only return intermediate features
        Returns:

        )r  zOutput shape must be NCHW.r   N)
r   rÚ   rÜ   rÙ   r*   r»   r¼   Ú	enumeraterP   rµ   )r-   r2   r	  rP   r
  r  r  ÚintermediatesÚtake_indicesÚ	max_indexÚlast_idxrÜ   Úfeat_idxrç   Úx_inters                  r0   Úforward_intermediatesz!SwiftFormer.forward_intermediatesÙ  sC  € ð* ˜YÐ&Ð&Ð&Ð(DÑ&Ô&Ð&ØˆÝ"6µs¸4¼;Ñ7GÔ7GÈÑ"QÔ"QÑˆ�iÝ�t”{Ñ#Ô# aÑ'ˆð �IŠI�a‰LŒLˆÝŒ9×!Ò!Ñ#Ô#ð 	1¨:ð 	1Ø”[ˆFˆFà”[  )¨a¡- Ô0ˆFå(¨Ñ0Ô0ð 	.ð 	.‰OˆH�eØ��a‘”ˆAØ˜<Ð'Ð'Øð  ˜H¨Ò0Ð0Ø"Ÿiši¨™lœl�G�Gà�GØ×$Ò$ WÑ-Ô-Ð-øàð 	!Ø Ð à�xÒÐØ—	’	˜!‘”ˆAà�-ÐÐr1   Ú
prune_normÚ
prune_headc                 óâ   — t          t          | j        ¦  «        |¦  «        \  }}| j        d|dz   …         | _        |rt          j        ¦   «         | _        |r|                      dd¦  «         |S )z@ Prune layers not required for specified intermediates.
        Nr   r   Ú )r   rÚ   rÜ   r(   rO   rP   r  )r-   r	  r  r  r  r  s         r0   Úprune_intermediate_layersz%SwiftFormer.prune_intermediate_layers  sq   € õ #7µs¸4¼;Ñ7GÔ7GÈÑ"QÔ"QÑˆ�iØ”k . 9¨q¡= .Ô1ˆŒØð 	&Ýœ™œˆDŒIØð 	)Ø×!Ò! ! RÑ(Ô(Ð(ØÐr1   c                 ó„   — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|S r5   )rÙ   rÜ   rP   r7   s     r0   Úforward_featureszSwiftFormer.forward_features  s4   € Ø�IŠI�a‰LŒLˆØ�KŠK˜‰NŒNˆØ�IŠI�a‰LŒLˆØˆr1   Ú
pre_logitsc                 ó:  — | j         dk    r|                     d¬¦  «        }|                      |¦  «        }|r|S |                      |¦  «        |                      |¦  «        }}| j        r)| j        r"t          j         	                    ¦   «         s||fS ||z   dz  S )NrÇ   )r_   rD   r‰   r_   )
rÐ   Úmeanrß   rà   rá   râ   r  r*   r»   r¼   )r-   r2   r  Úx_dists       r0   Úforward_headzSwiftFormer.forward_head!  s¤   € ØÔ˜uÒ$Ð$Ø—’˜6�Ñ"Ô"ˆAØ�NŠN˜1ÑÔˆØð 	ØˆHØ—I’I˜a‘L”L $§.¢.°Ñ"3Ô"3ˆ6ˆØÔ"ð 	$ t¤}ð 	$½U¼Y×=SÒ=SÑ=UÔ=Uð 	$à�f�9Ðð ˜‘J !Ñ#Ð#r1   c                 óZ   — |                       |¦  «        }|                      |¦  «        }|S r5   )r  r!  r7   s     r0   r8   zSwiftFormer.forward/  s-   € Ø×!Ò! !Ñ$Ô$ˆØ×Ò˜aÑ Ô ˆØˆr1   ©F)Tr5   )NFFr  F)r   FT)$r9   r:   r;   r(   rl   r   r<   r>   r	   rT   r=   Ústrr'   rã   r*   r»   Úignorer   rö   r   r   rú   rþ   r   r  r   r  r  r?   r
   r  r  r  r!  r8   r@   rA   s   @r0   r   r   X  sû  ø€ € € € € ð !-  Ø$6Ð$6Ð$6ØØ&?Ð&?Ð&?Ø)+¬Ø#$Ø ØØ#Ø!Ø$&Ø$(Ø,0Ø$Ø!#ØØØð'L#ð L#à˜”IðL#ð ˜Sœ	ðL#ð ð	L#ð
 ˜dœðL#ð ˜BœI”ðL#ð !ðL#ð ðL#ð ðL#ð ðL#ð ðL#ð "ðL#ð "ðL#ð %*ðL#ð ðL#ð  ð!L#ð" ð#L#ð L#ð L#ð L#ð L#ð L#ð\	1ð 	1ð 	1ð „YÔð ð ð ð ñ Ôðð „YÔð	ð 	 Dð 	°T¸#¸s¸(´^ð 	ð 	ð 	ñ Ôð	ð „YÔð*ð *¨Tð *ð *ð *ñ Ôð*ð „YÔð)  b¤i°´Ð&:Ô ;ð )ð )ð )ñ Ôð)ð,ð ,¨Cð ,¸hÀs¼mð ,ð ,ð ,ð ,ð „YÔð)ð )¨Tð )ð )ð )ñ Ôð)ð 8<ØØ$Ø$Ø',ð0 ð 0 àŒ|ð0 ð ˜e C¨¨c¬ NÔ3Ô4ð0 ð ð	0 ð
 ð0 ð ð0 ð !%ð0 ð 
ˆt�E”LÔ! 5¨¬°t¸E¼LÔ7IÐ)IÔ#JÐJÔ	Kð0 ð 0 ð 0 ð 0 ðh ./Ø$Ø#ð	ð à˜3  S¤	˜>Ô*ðð ðð ð	ð ð ð ð  %¤,ð °5´<ð ð ð ð ð$ð $˜eœlð $¸ð $ð $ð $ð $ð˜œð ð ð ð ð ð ð ð r1   Ú
state_dictÚmodelr3   c                 óh  — |                       d| ¦  «        } d| v r| S i }|                      ¦   «         D ]þ\  }}|                     dd¦  «        }|                     dd¦  «        }|                     dd¦  «        }|                     d	d
¦  «        }|                     dd¦  «        }t          j        dd|¦  «        }t          j        d|¦  «        }|rYt          |                     d¦  «        ¦  «        |                     d¦  «        }}|dz  }|dz  dk    r	d|› d|› �}nd|dz   › d|› �}|||<   Œÿ|S )Nr'  zstem.0.weightzpatch_embed.zstem.z
dist_head.z
head_dist.z
attn.Proj.z
attn.proj.z.layer_scale_1z.layer_scale_1.gammaz.layer_scale_2z.layer_scale_2.gammaz\.layer_scale(?=$|\.)z.layer_scale.gammaz^network\.(\d+)\.(.*)r   r_   r   rÓ   z.blocks.z.downsample.)ÚgetÚitemsÚreplaceÚreÚsubÚmatchr<   Úgroup)	r&  r'  Úout_dictÚkÚvrò   Ún_idxÚrestÚ	stage_idxs	            r0   Úcheckpoint_filter_fnr6  5  sU  € Ø—’ ¨Ñ4Ô4€JØ˜*Ð$Ð$ØÐà€HØ× Ò Ñ"Ô"ð ð ‰ˆˆ1Ø�IŠI�n gÑ.Ô.ˆØ�IŠI�l LÑ1Ô1ˆØ�IŠI�l LÑ1Ô1ˆØ�IŠIÐ&Ð(>Ñ?Ô?ˆØ�IŠIÐ&Ð(>Ñ?Ô?ˆÝŒFÐ+Ð-AÀ1ÑEÔEˆÝŒHÐ-¨qÑ1Ô1ˆØð 	>Ý˜aŸgšg a™jœj™/œ/¨1¯7ª7°1©:¬:�4ˆEØ ™
ˆIØ�q‰y˜AŠ~ˆ~Ø7˜iÐ7Ð7°Ð7Ð7��à=˜i¨™kÐ=Ð=°tÐ=Ð=�àˆ�‰ˆØ€Or1   r  Úurlrä   c                 ó>   — | ddd dddt           t          dddd	d
ddœ|¥S )NrÆ   )rD   éà   r9  Tgffffffî?Úbicubiczstem.0r   z
apache-2.0zarXiv:2303.15446zdSwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision Applicationsz'https://github.com/Amshaker/SwiftFormer)r7  rÏ   Ú
input_sizeÚ	pool_sizeÚfixed_input_sizeÚcrop_pctÚinterpolationr  rê   Ú
first_convÚ
classifierÚlicenseÚ	paper_idsÚ
paper_nameÚ
origin_urlr   )r7  rä   s     r0   Ú_cfgrF  O  sE   € àØ¨=ÀtÐaeØ¨)Ý%Õ.BØÐ.CØØ'Ø|Ø?ðð ð ðð r1   ztimm/)Ú	hf_hub_id)zswiftformer_xs.dist_in1kzswiftformer_s.dist_in1kzswiftformer_l1.dist_in1kzswiftformer_l3.dist_in1kFÚvariantÚ
pretrainedc                 ó\   — t          t          | |ft          t          dd¬¦  «        dœ|¤Ž}|S )N)r   r   r_   rD   T)Úout_indicesÚflatten_sequential)Úpretrained_filter_fnÚfeature_cfg)r   r   r6  rÛ   )rH  rI  rä   r'  s       r0   Ú_create_swiftformerrO  n  sF   € Ý Ý�W˜jðå1Ý \ÀdÐKÑKÔKðð ð ð	ð €Eð €Lr1   c           	      ó^   — t          g d¢g d¢¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )Nr¾   rÁ   ©r¯   rÉ   Úswiftformer_xsrI  )rR  ©rÛ   rO  ©rI  rä   Ú
model_argss      r0   rR  rR  x  sI   € å˜\˜\˜\Ð6HÐ6HÐ6HÐIÑIÔI€JÝÐeÐe¸JÐeÍ$ÈzÐJdÐJdÐ]cÐJdÐJdÐeÐeÐer1   c           	      ó^   — t          g d¢g d¢¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )N)rD   rD   é	   r¿   )rÂ   rW   é¨   r9  rQ  Úswiftformer_srI  )rY  rS  rT  s      r0   rY  rY  ~  sI   € å˜\˜\˜\Ð6HÐ6HÐ6HÐIÑIÔI€JÝÐdÐd¸:ÐdÍÈjÐIcÐIcÐ\bÐIcÐIcÐdÐdÐdr1   c           	      ó^   — t          g d¢g d¢¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )N)rÀ   rD   é
   é   )rÂ   é`   éÀ   i€  rQ  Úswiftformer_l1rI  )r_  rS  rT  s      r0   r_  r_  ƒ  sI   € å˜]˜]˜]Ð7IÐ7IÐ7IÐJÑJÔJ€JÝÐeÐe¸JÐeÍ$ÈzÐJdÐJdÐ]cÐJdÐJdÐeÐeÐer1   c           	      ó^   — t          g d¢g d¢¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )N)rÀ   rÀ   é   r¿   )rW   é€   i@  r{   rQ  Úswiftformer_l3rI  )rc  rS  rT  s      r0   rc  rc  ‰  sI   € å˜]˜]˜]Ð7JÐ7JÐ7JÐKÑKÔK€JÝÐeÐe¸JÐeÍ$ÈzÐJdÐJdÐ]cÐJdÐJdÐeÐeÐer1   )r  r#  ):rR   r,  Útypingr   r   r   r   r   r   r	   r
   r*   Útorch.nnr(   Útorch.nn.functionalÚ
functionalrŽ   Ú	timm.datar   r   Útimm.layersr   r   r   r   r   r   Ú_builderr   Ú	_featuresr   Ú_manipulater   Ú	_registryr   r   Ú__all__rT   r   rC   rV   rn   rz   r›   r    r­   r   r$  r?   r6  rF  Údefault_cfgsr>   rO  rR  rY  r_  rc  r¦   r1   r0   ú<module>rp     s  ððð ð 
€	€	€	Ø EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ EÐ Eà €€€Ø Ð Ð Ð Ð Ð Ø Ð Ð Ð Ð Ð Ð Ð Ð à AÐ AÐ AÐ AÐ AÐ AÐ AÐ AØ _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ð _Ø *Ð *Ð *Ð *Ð *Ð *Ø +Ð +Ð +Ð +Ð +Ð +Ø 'Ð 'Ð 'Ð 'Ð 'Ð 'Ø <Ð <Ð <Ð <Ð <Ð <Ð <Ð <àˆ/€ð	Fð 	Fð 	Fð 	Fð 	F�2”9ñ 	Fô 	Fð 	Fðð ð ð ð �”	ñ ô ð ð>%ð %ð %ð %ð %�"”)ñ %ô %ð %ðP"ð "ð "ð "ð "ˆ"Œ)ñ "ô "ð "ðJ%ð %ð %ð %ð % ¤ñ %ô %ð %ðP$ð $ð $ð $ð $˜"œ)ñ $ô $ð $ðN0ð 0ð 0ð 0ð 0ˆBŒIñ 0ô 0ð 0ðf=ð =ð =ð =ð =ˆBŒIñ =ô =ð =ð@Zð Zð Zð Zð Z�"”)ñ Zô Zð Zðz T¨#¨u¬|Ð*;Ô%<ð ÀRÄYð ÐSWÐX[Ð]bÔ]iÐXiÔSjð ð ð ð ð4ð ˆcð  #ð ¨$¨s°C¨x¬.ð ð ð ð ð %Ð$Ø $ Øð!ñ !ô !ð  $˜tØð ñ  ô  ð !% Øð!ñ !ô !ð !% Øð!ñ !ô !ð&ð &ñ ô €ð ð  ð °$ð È#ð ÐR]ð ð ð ð ð ðfð f˜tð f°sð f¸{ð fð fð fñ „ðfð
 ðeð e˜dð e°cð e¸kð eð eð eñ „ðeð ðfð f˜tð f°sð f¸{ð fð fð fñ „ðfð
 ðfð f˜tð f°sð f¸{ð fð fð fñ „ðfð fð fr1   