§
    ·¯¸j×´  ã                   ó
  — d Z ddlmZ ddlmZ ddl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ZmZ ddlmZmZmZmZmZmZmZmZmZ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+m,Z, ddl-m.Z.m/Z/m0Z0 dgZ1 G d„ dej2        ¦  «        Z3 G d„ dej2        ¦  «        Z4 G d„ dej2        ¦  «        Z5 G d„ dej2        ¦  «        Z6 G d„ dej2        ¦  «        Z7 G d„ dej2        ¦  «        Z8de9de:fd„Z;ddd e eed!¬"¦  «        ddfd#e<d$e<de9d%e:d&ed'edej=        fd(„Z> G d)„ dej2        ¦  «        Z?dzd*ej2        d+e9d,e:ddfd-„Z@ ejA        ¦   «         d{d/ej2        d0e9d1e9fd2„¦   «         ZBd|d4e9d5e:d6ede?fd7„ZCd|d4e9d5e:d6ede?fd8„ZDd}d9e9d6edee9ef         fd:„ZE e.i d; eEd<d=d ¬>¦  «        “d? eEd<d=d ¬>¦  «        “d@ eEd<dAdBdCd=d ¬D¦  «        “dE eEd<dFdGdCd ¬H¦  «        “dI eEd<dFdGdCd ¬H¦  «        “dJ eEd<dFdGdCd ¬H¦  «        “dK eEd<dFdGdCd ¬H¦  «        “dL eEd<dFdGdCd ¬H¦  «        “dM eEd<dNdOdCd ¬H¦  «        “dP eEd<dQd ¬R¦  «        “dS eEd<dQd ¬R¦  «        “dT eEd<dQd ¬R¦  «        “dU eEd<dQd ¬R¦  «        “dV eEd<dQd ¬R¦  «        “dW eEd<dQd ¬R¦  «        “dX eEd<d=dYdZdC¬[¦  «        “d\ eEd<d=dYdZdCd]¬^¦  «        “ eEd<d=dYdZdC¬[¦  «         eEd<d=dYdZdCd]¬^¦  «         eEd<dCdAdBdFd=d]¬_¦  «         eEd<d=d`dZdC¬[¦  «         eEd=d]¬a¦  «         eEd=d]¬a¦  «         eEd<d=d`dZdC¬[¦  «         eEd=d]¬a¦  «         eEd=¬b¦  «         eEd=d]¬a¦  «         eEd<d=d]d`dZdC¬c¦  «         eEd<d=d]d`dZdC¬c¦  «         eEd=d]¬a¦  «        ddœ¥¦  «        ZFe/d|d5e:d6ede?fde„¦   «         ZGe/d|d5e:d6ede?fdf„¦   «         ZHe/d|d5e:d6ede?fdg„¦   «         ZIe/d|d5e:d6ede?fdh„¦   «         ZJe/d|d5e:d6ede?fdi„¦   «         ZKe/d|d5e:d6ede?fdj„¦   «         ZLe/d|d5e:d6ede?fdk„¦   «         ZMe/d|d5e:d6ede?fdl„¦   «         ZNe/d|d5e:d6ede?fdm„¦   «         ZOe/d|d5e:d6ede?fdn„¦   «         ZPe/d|d5e:d6ede?fdo„¦   «         ZQe/d|d5e:d6ede?fdp„¦   «         ZRe/d|d5e:d6ede?fdq„¦   «         ZSe/d|d5e:d6ede?fdr„¦   «         ZTe/d|d5e:d6ede?fds„¦   «         ZUe/d|d5e:d6ede?fdt„¦   «         ZVe/d|d5e:d6ede?fdu„¦   «         ZWe/d|d5e:d6ede?fdv„¦   «         ZXe/d|d5e:d6ede?fdw„¦   «         ZYe/d|d5e:d6ede?fdx„¦   «         ZZ e0e[dEdIdJdKdLdMdPdSdTdUdVdWd;d?d@dyœ¦  «         dS )~a/  Pre-Activation ResNet v2 with GroupNorm and Weight Standardization.

A PyTorch implementation of ResNetV2 adapted from the Google Big-Transfer (BiT) source code
at https://github.com/google-research/big_transfer to match timm interfaces. The BiT weights have
been included here as pretrained models from their original .NPZ checkpoints.

Additionally, supports non pre-activation bottleneck for use as a backbone for Vision Transformers (ViT) and
extra padding support to allow porting of official Hybrid ResNet pretrained weights from
https://github.com/google-research/vision_transformer

Thanks to the Google team for the above two repositories and associated papers:
* Big Transfer (BiT): General Visual Representation Learning - https://arxiv.org/abs/1912.11370
* An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale - https://arxiv.org/abs/2010.11929
* Knowledge distillation: A good teacher is patient and consistent - https://arxiv.org/abs/2106.05237

Original copyright of Google code below, modifications by Ross Wightman, Copyright 2020.
é    )ÚOrderedDict)Úpartial)ÚAnyÚCallableÚDictÚListÚOptionalÚTupleÚUnionN©ÚIMAGENET_INCEPTION_MEANÚIMAGENET_INCEPTION_STD)ÚGroupNormActÚBatchNormAct2dÚEvoNorm2dS0ÚFilterResponseNormTlu2dÚClassifierHeadÚDropPathÚcalculate_drop_path_ratesÚAvgPool2dSameÚcreate_pool2dÚ	StdConv2dÚcreate_conv2dÚget_act_layerÚget_norm_act_layerÚmake_divisibleÚget_device_dtypeé   )Úbuild_model_with_cfg)Úfeature_take_indices)Ú_load_npz_checkpoint)Úcheckpoint_seqÚnamed_applyÚadapt_input_conv)Úgenerate_default_cfgsÚregister_modelÚregister_model_deprecationsÚResNetV2c                   óä   ‡ — e Zd ZdZ	 	 	 	 	 	 	 	 	 	 	 	 	 ddedee         ded	ed
edee         dedee         dee         dee         dee         defˆ fd„Zdd„Z	de
j        de
j        fd„Zˆ xZS )ÚPreActBasiczAPre-activation basic block (not in typical 'v2' implementations).Nç      ð?r   ç        Úin_chsÚout_chsÚbottle_ratioÚstrideÚdilationÚfirst_dilationÚgroupsÚ	act_layerÚ
conv_layerÚ
norm_layerÚ
proj_layerÚdrop_path_ratec           
      ó  •— ||dœ}t          ¦   «                              ¦   «          |p|}|	pt          }	|
pt          t          d¬¦  «        }
|p|}t          ||z  ¦  «        }|�)|dk    s||k    s||k    r |||f|||d|	|
dœ|¤Ž| _        nd| _         |
|fi |¤Ž| _         |	||df|||d	œ|¤Ž| _         |
|fi |¤Ž| _	         |	||df||d
œ|¤Ž| _
        |dk    rt          |¦  «        nt          j        ¦   «         | _        dS )aw  Initialize PreActBasic block.

        Args:
            in_chs: Input channels.
            out_chs: Output channels.
            bottle_ratio: Bottleneck ratio (not used in basic block).
            stride: Stride for convolution.
            dilation: Dilation rate.
            first_dilation: First dilation rate.
            groups: Group convolution size.
            act_layer: Activation layer type.
            conv_layer: Convolution layer type.
            norm_layer: Normalization layer type.
            proj_layer: Projection/downsampling layer type.
            drop_path_rate: Stochastic depth drop rate.
        ©ÚdeviceÚdtypeé    ©Ú
num_groupsNr   T©r0   r1   r2   Úpreactr5   r6   é   ©r0   r1   r3   )r1   r3   r   )ÚsuperÚ__init__r   r   r   r   Ú
downsampleÚnorm1Úconv1Únorm2Úconv2r   ÚnnÚIdentityÚ	drop_path©Úselfr-   r.   r/   r0   r1   r2   r3   r4   r5   r6   r7   r8   r;   r<   ÚddÚmid_chsÚ	__class__s                    €úd/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/timm/models/resnetv2.pyrE   zPreActBasic.__init__6   s‡  ø€ ðB ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ'Ð3¨8ˆØÐ,¥9ˆ
ØÐG¥7­<ÀBÐ#GÑ#GÔ#Gˆ
ØÐ#˜VˆÝ  ¨<Ñ!7Ñ8Ô8ˆàÐ! v°¢{ {°nÈÒ6PÐ6PÐTZÐ^eÒTeÐTeØ(˜jØØð
ð Ø!Ø-ØØ%Ø%ð
ð 
ð ð
ð 
ˆDŒOˆOð #ˆDŒOà�Z Ð-Ð-¨"Ð-Ð-ˆŒ
Ø�Z ¨°Ðp¸6ÈNÐciÐpÐpÐmoÐpÐpˆŒ
Ø�Z Ð.Ð.¨2Ð.Ð.ˆŒ
Ø�Z ¨°!Ð\¸hÈvÐ\Ð\ÐY[Ð\Ð\ˆŒ
Ø5CÀaÒ5GÐ5G� .Ñ1Ô1Ð1ÍRÌ[É]Ì]ˆŒˆˆó    Úreturnc                 óX   — t           j                             | j        j        ¦  «         dS )zLZero-initialize the last convolution weight (not applicable to basic block).N)rK   ÚinitÚzeros_rJ   Úweight©rO   s    rS   Úzero_init_lastzPreActBasic.zero_init_lastt   ó!   € å
Œ�Š�t”zÔ(Ñ)Ô)Ð)Ð)Ð)rT   Úxc                 ó  — |                       |¦  «        }|}| j        �|                      |¦  «        }|                      |¦  «        }|                      |                      |¦  «        ¦  «        }|                      |¦  «        }||z   S ©zoForward pass.

        Args:
            x: Input tensor.

        Returns:
            Output tensor.
        )rG   rF   rH   rJ   rI   rM   ©rO   r]   Úx_preactÚshortcuts       rS   ÚforwardzPreActBasic.forwardx   s|   € ð —:’:˜a‘=”=ˆð ˆØŒ?Ð&Ø—’ xÑ0Ô0ˆHð �JŠJ�xÑ Ô ˆØ�JŠJ�t—z’z !‘}”}Ñ%Ô%ˆØ�NŠN˜1ÑÔˆØ�8‰|ÐrT   )Nr+   r   r   Nr   NNNNr,   NN©rU   N©Ú__name__Ú
__module__Ú__qualname__Ú__doc__Úintr	   Úfloatr   rE   r[   ÚtorchÚTensorrc   Ú__classcell__©rR   s   @rS   r*   r*   3   sQ  ø€ € € € € ØKÐKð
 &*Ø"%ØØØ,0ØØ,0Ø-1Ø-1Ø-1Ø$&ØØð<[ð <[àð<[ð ˜c”]ð<[ð  ð	<[ð
 ð<[ð ð<[ð % SœMð<[ð ð<[ð   Ô)ð<[ð ! Ô*ð<[ð ! Ô*ð<[ð ! Ô*ð<[ð "ð<[ð <[ð <[ð <[ð <[ð <[ð|*ð *ð *ð *ð˜œð ¨%¬,ð ð ð ð ð ð ð ð rT   r*   c                   óä   ‡ — e Zd ZdZ	 	 	 	 	 	 	 	 	 	 	 	 	 ddedee         ded	ed
edee         dedee         dee         dee         dee         defˆ fd„Zdd„Z	de
j        de
j        fd„Zˆ xZS )ÚPreActBottlenecka  Pre-activation (v2) bottleneck block.

    Follows the implementation of "Identity Mappings in Deep Residual Networks":
    https://github.com/KaimingHe/resnet-1k-layers/blob/master/resnet-pre-act.lua

    Except it puts the stride on 3x3 conv when available.
    Nç      Ð?r   r,   r-   r.   r/   r0   r1   r2   r3   r4   r5   r6   r7   r8   c           
      ó  •— ||dœ}t          ¦   «                              ¦   «          |p|}|	pt          }	|
pt          t          d¬¦  «        }
|p|}t          ||z  ¦  «        }|� |||f|||d|	|
dœ|¤Ž| _        nd| _         |
|fi |¤Ž| _         |	||dfi |¤Ž| _         |
|fi |¤Ž| _	         |	||df|||d	œ|¤Ž| _
         |
|fi |¤Ž| _         |	||dfi |¤Ž| _        |d
k    rt          |¦  «        nt          j        ¦   «         | _        dS )ab  Initialize PreActBottleneck block.

        Args:
            in_chs: Input channels.
            out_chs: Output channels.
            bottle_ratio: Bottleneck ratio.
            stride: Stride for convolution.
            dilation: Dilation rate.
            first_dilation: First dilation rate.
            groups: Group convolution size.
            act_layer: Activation layer type.
            conv_layer: Convolution layer type.
            norm_layer: Normalization layer type.
            proj_layer: Projection/downsampling layer type.
            drop_path_rate: Stochastic depth drop rate.
        r:   r=   r>   NTr@   r   rB   rC   r   )rD   rE   r   r   r   r   rF   rG   rH   rI   rJ   Únorm3Úconv3r   rK   rL   rM   rN   s                    €rS   rE   zPreActBottleneck.__init__˜   s—  ø€ ðB ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ'Ð3¨8ˆØÐ,¥9ˆ
ØÐG¥7­<ÀBÐ#GÑ#GÔ#Gˆ
ØÐ#˜VˆÝ  ¨<Ñ!7Ñ8Ô8ˆàÐ!Ø(˜jØØð
ð Ø!Ø-ØØ%Ø%ð
ð 
ð ð
ð 
ˆDŒOˆOð #ˆDŒOà�Z Ð-Ð-¨"Ð-Ð-ˆŒ
Ø�Z ¨°Ð9Ð9°bÐ9Ð9ˆŒ
Ø�Z Ð.Ð.¨2Ð.Ð.ˆŒ
Ø�Z ¨°!Ðq¸FÈ^ÐdjÐqÐqÐnpÐqÐqˆŒ
Ø�Z Ð.Ð.¨2Ð.Ð.ˆŒ
Ø�Z ¨°!Ð:Ð:°rÐ:Ð:ˆŒ
Ø5CÀaÒ5GÐ5G� .Ñ1Ô1Ð1ÍRÌ[É]Ì]ˆŒˆˆrT   rU   c                 óX   — t           j                             | j        j        ¦  «         dS )z,Zero-initialize the last convolution weight.N)rK   rW   rX   ru   rY   rZ   s    rS   r[   zPreActBottleneck.zero_init_lastØ   r\   rT   r]   c                 óf  — |                       |¦  «        }|}| j        �|                      |¦  «        }|                      |¦  «        }|                      |                      |¦  «        ¦  «        }|                      |                      |¦  «        ¦  «        }|                      |¦  «        }||z   S r_   )rG   rF   rH   rJ   rI   ru   rt   rM   r`   s       rS   rc   zPreActBottleneck.forwardÜ   s—   € ð —:’:˜a‘=”=ˆð ˆØŒ?Ð&Ø—’ xÑ0Ô0ˆHð �JŠJ�xÑ Ô ˆØ�JŠJ�t—z’z !‘}”}Ñ%Ô%ˆØ�JŠJ�t—z’z !‘}”}Ñ%Ô%ˆØ�NŠN˜1ÑÔˆØ�8‰|ÐrT   ©Nrr   r   r   Nr   NNNNr,   NNrd   re   ro   s   @rS   rq   rq   �   sU  ø€ € € € € ðð ð &*Ø"&ØØØ,0ØØ,0Ø-1Ø-1Ø-1Ø$&ØØð>[ð >[àð>[ð ˜c”]ð>[ð  ð	>[ð
 ð>[ð ð>[ð % SœMð>[ð ð>[ð   Ô)ð>[ð ! Ô*ð>[ð ! Ô*ð>[ð ! Ô*ð>[ð "ð>[ð >[ð >[ð >[ð >[ð >[ð@*ð *ð *ð *ð˜œð ¨%¬,ð ð ð ð ð ð ð ð rT   rq   c                   óä   ‡ — e Zd ZdZ	 	 	 	 	 	 	 	 	 	 	 	 	 ddedee         ded	ed
edee         dedee         dee         dee         dee         defˆ fd„Zdd„Z	de
j        de
j        fd„Zˆ xZS )Ú
BottleneckzUNon Pre-activation bottleneck block, equiv to V1.5/V1b Bottleneck. Used for ViT.
    Nrr   r   r,   r-   r.   r/   r0   r1   r2   r3   r4   r5   r6   r7   r8   c           	      ó^  •— ||dœ}t          ¦   «                              ¦   «          |p|}|pt          j        }|	pt          }	|
pt          t          d¬¦  «        }
|p|}t          ||z  ¦  «        }|� |||f||d|	|
dœ|¤Ž| _        nd | _         |	||dfi |¤Ž| _	         |
|fi |¤Ž| _
         |	||df|||dœ|¤Ž| _         |
|fi |¤Ž| _         |	||dfi |¤Ž| _         |
|fd	di|¤Ž| _        |d
k    rt          |¦  «        nt          j        ¦   «         | _         |d¬¦  «        | _        d S )Nr:   r=   r>   F)r0   r1   rA   r5   r6   r   rB   rC   Ú	apply_actr   T)Úinplace)rD   rE   rK   ÚReLUr   r   r   r   rF   rH   rG   rJ   rI   ru   rt   r   rL   rM   Úact3rN   s                    €rS   rE   zBottleneck.__init__÷   s¸  ø€ ð" ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ'Ð3¨8ˆØÐ(¥¤ˆ	ØÐ,¥9ˆ
ØÐG¥7­<ÀBÐ#GÑ#GÔ#Gˆ
ØÐ#˜VˆÝ  ¨<Ñ!7Ñ8Ô8ˆàÐ!Ø(˜jØØð	ð Ø!ØØ%Ø%ð	ð 	ð ð	ð 	ˆDŒOˆOð #ˆDŒOà�Z ¨°Ð9Ð9°bÐ9Ð9ˆŒ
Ø�Z Ð.Ð.¨2Ð.Ð.ˆŒ
Ø�Z ¨°!Ðq¸FÈ^ÐdjÐqÐqÐnpÐqÐqˆŒ
Ø�Z Ð.Ð.¨2Ð.Ð.ˆŒ
Ø�Z ¨°!Ð:Ð:°rÐ:Ð:ˆŒ
Ø�Z Ð?Ð?°5Ð?¸BÐ?Ð?ˆŒ
Ø5CÀaÒ5GÐ5G� .Ñ1Ô1Ð1ÍRÌ[É]Ì]ˆŒØ�I dÐ+Ñ+Ô+ˆŒ	ˆ	ˆ	rT   rU   c                 óˆ   — t          | j        dd¦  «        �+t          j                             | j        j        ¦  «         dS dS )z+Zero-initialize the last batch norm weight.rY   N)Úgetattrrt   rK   rW   rX   rY   rZ   s    rS   r[   zBottleneck.zero_init_last(  s=   € å�4”:˜x¨Ñ.Ô.Ð:ÝŒG�NŠN˜4œ:Ô,Ñ-Ô-Ð-Ð-Ð-ð ;Ð:rT   r]   c                 ó˜  — |}| j         �|                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      ||z   ¦  «        }|S r_   )	rF   rH   rG   rJ   rI   ru   rt   rM   r   )rO   r]   rb   s      rS   rc   zBottleneck.forward-  s©   € ð ˆØŒ?Ð&Ø—’ qÑ)Ô)ˆHð �JŠJ�q‰MŒMˆØ�JŠJ�q‰MŒMˆØ�JŠJ�q‰MŒMˆØ�JŠJ�q‰MŒMˆØ�JŠJ�q‰MŒMˆØ�JŠJ�q‰MŒMˆØ�NŠN˜1ÑÔˆØ�IŠI�a˜(‘lÑ#Ô#ˆØˆrT   rx   rd   re   ro   s   @rS   rz   rz   ô   sB  ø€ € € € € ðð ð
 &*Ø"&ØØØ,0ØØ,0Ø-1Ø-1Ø-1Ø$&ØØð/,ð /,àð/,ð ˜c”]ð/,ð  ð	/,ð
 ð/,ð ð/,ð % SœMð/,ð ð/,ð   Ô)ð/,ð ! Ô*ð/,ð ! Ô*ð/,ð ! Ô*ð/,ð "ð/,ð /,ð /,ð /,ð /,ð /,ðb.ð .ð .ð .ð
˜œð ¨%¬,ð ð ð ð ð ð ð ð rT   rz   c                   óž   ‡ — e Zd ZdZ	 	 	 	 	 	 	 	 ddedededed	ee         d
edee         dee         fˆ fd„Zde	j
        de	j
        fd„Zˆ xZS )ÚDownsampleConvz$1x1 convolution downsampling module.r   NTr-   r.   r0   r1   r2   rA   r5   r6   c                 óÀ   •— |	|
dœ}t          ¦   «                              ¦   «           |||dfd|i|¤Ž| _        |rt          j        ¦   «         n
 ||fddi|¤Ž| _        d S )Nr:   r   r0   r|   F)rD   rE   ÚconvrK   rL   Únorm)rO   r-   r.   r0   r1   r2   rA   r5   r6   r;   r<   rP   rR   s               €rS   rE   zDownsampleConv.__init__J  s|   ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ�J˜v w°ÐGÐG¸&ÐGÀBÐGÐGˆŒ	Ø%+Ð[•B”K‘M”M�M°°¸GÐ1[Ð1[ÈuÐ1[ÐXZÐ1[Ð1[ˆŒ	ˆ	ˆ	rT   r]   rU   c                 óR   — |                       |                      |¦  «        ¦  «        S ©ztForward pass.

        Args:
            x: Input tensor.

        Returns:
            Downsampled tensor.
        )r‡   r†   ©rO   r]   s     rS   rc   zDownsampleConv.forward\  s    € ð �yŠy˜Ÿš 1™œÑ&Ô&Ð&rT   ©r   r   NTNNNN©rf   rg   rh   ri   rj   r	   Úboolr   rE   rl   rm   rc   rn   ro   s   @rS   r„   r„   G  só   ø€ € € € € Ø.Ð.ð ØØ,0ØØ-1Ø-1ØØð\ð \àð\ð ð\ð ð	\ð
 ð\ð % SœMð\ð ð\ð ! Ô*ð\ð ! Ô*ð\ð \ð \ð \ð \ð \ð$	'˜œð 	'¨%¬,ð 	'ð 	'ð 	'ð 	'ð 	'ð 	'ð 	'ð 	'rT   r„   c                   óž   ‡ — e Zd ZdZ	 	 	 	 	 	 	 	 ddedededed	ee         d
edee         dee         fˆ fd„Zde	j
        de	j
        fd„Zˆ xZS )ÚDownsampleAvgz/AvgPool downsampling as in 'D' ResNet variants.r   NTr-   r.   r0   r1   r2   rA   r5   r6   c                 ó„  •— |	|
dœ}t          ¦   «                              ¦   «          |dk    r|nd}|dk    s|dk    r4|dk    r|dk    rt          nt          j        } |d|dd¬¦  «        | _        nt          j        ¦   «         | _         |||dfddi|¤Ž| _        |rt          j        ¦   «         n
 ||fddi|¤Ž| _        d S )	Nr:   r   é   TF)Ú	ceil_modeÚcount_include_padr0   r|   )	rD   rE   r   rK   Ú	AvgPool2dÚpoolrL   r†   r‡   )rO   r-   r.   r0   r1   r2   rA   r5   r6   r;   r<   rP   Ú
avg_strideÚavg_pool_fnrR   s                 €rS   rE   zDownsampleAvg.__init__k  så   ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐØ'¨1š}˜}�V�V°!ˆ
Ø�AŠ:ˆ:˜ Aš˜Ø+5¸ª?¨?¸xÈ!º|¸|�-˜-ÕQSÔQ]ˆKØ#˜ A z¸TÐUZÐ[Ñ[Ô[ˆDŒIˆIåœ™œˆDŒIØ�J˜v w°ÐBÐB¸!ÐB¸rÐBÐBˆŒ	Ø%+Ð[•B”K‘M”M�M°°¸GÐ1[Ð1[ÈuÐ1[ÐXZÐ1[Ð1[ˆŒ	ˆ	ˆ	rT   r]   rU   c                 óx   — |                       |                      |                      |¦  «        ¦  «        ¦  «        S r‰   )r‡   r†   r•   rŠ   s     rS   rc   zDownsampleAvg.forwardƒ  s,   € ð �yŠy˜Ÿš 4§9¢9¨Q¡<¤<Ñ0Ô0Ñ1Ô1Ð1rT   r‹   rŒ   ro   s   @rS   r�   r�   h  só   ø€ € € € € Ø9Ð9ð ØØ,0ØØ-1Ø-1ØØð\ð \àð\ð ð\ð ð	\ð
 ð\ð % SœMð\ð ð\ð ! Ô*ð\ð ! Ô*ð\ð \ð \ð \ð \ð \ð0	2˜œð 	2¨%¬,ð 	2ð 	2ð 	2ð 	2ð 	2ð 	2ð 	2ð 	2rT   r�   c                   óÎ   ‡ — e Zd ZdZddddedddfdededed	ed
ededededee	e                  de
dee
         dee
         dee
         defˆ fd„Zdej        dej        fd„Zˆ xZS )ÚResNetStagezResNet Stage.rr   r   FNr-   r.   r0   r1   Údepthr/   r3   Úavg_downÚ	block_dprÚblock_fnr4   r5   r6   Úblock_kwargsc                 ó®  •— t          ¦   «                              ¦   «          d| _        |dv rdnd}t          |||¬¦  «        }|rt          nt
          }|}t          j        ¦   «         | _        t          |¦  «        D ]X}|	r|	|         nd}|dk    r|nd}| j         
                    t          |¦  «         |
||f|||||||dœ|¤|¤Ž¦  «         |}|}d }ŒYd S )	NF)r   r‘   r   r‘   )r4   r5   r6   r,   r   )r0   r1   r/   r3   r2   r7   r8   )rD   rE   Úgrad_checkpointingÚdictr�   r„   rK   Ú
SequentialÚblocksÚrangeÚ
add_moduleÚstr)rO   r-   r.   r0   r1   r›   r/   r3   rœ   r�   rž   r4   r5   r6   rŸ   r2   Úlayer_kwargsr7   Úprev_chsÚ	block_idxr8   rR   s                        €rS   rE   zResNetStage.__init__‘  s-  ø€ õ" 	‰Œ×ÒÑÔÐØ"'ˆÔà&¨&Ð0Ð0˜˜°aˆÝ i¸JÐS]Ð^Ñ^Ô^ˆØ&.ÐB•]�]µNˆ
ØˆÝ”m‘o”oˆŒÝ˜u™œð 	ð 	ˆIØ5>ÐF˜Y yÔ1Ð1ÀBˆNØ(¨Aš~˜~�V�V°1ˆFØŒK×"Ò"¥3 y¡>¤>°8°8ØØð4ð Ø!Ø)ØØ-Ø%Ø-ð4ð 4ð ð4ð ð4ð 4ñ ô ð ð ˆHØ%ˆNØˆJˆJð%	ð 	rT   r]   rU   c                 ó¦   — | j         r4t          j                             ¦   «         st	          | j        |¦  «        }n|                      |¦  «        }|S )z�Forward pass through all blocks in the stage.

        Args:
            x: Input tensor.

        Returns:
            Output tensor.
        )r¡   rl   ÚjitÚis_scriptingr"   r¤   rŠ   s     rS   rc   zResNetStage.forward¾  sK   € ð Ô"ð 	­5¬9×+AÒ+AÑ+CÔ+Cð 	Ý˜tœ{¨AÑ.Ô.ˆAˆAà—’˜A‘”ˆAØˆrT   )rf   rg   rh   ri   rq   rj   rk   r�   r	   r   r   r   rE   rl   rm   rc   rn   ro   s   @rS   rš   rš   �  s+  ø€ € € € € ØÐð #'ØØ"Ø/3Ø!1Ø,0Ø-1Ø-1ð+ð +àð+ð ð+ð ð	+ð
 ð+ð ð+ð  ð+ð ð+ð ð+ð    U¤Ô,ð+ð ð+ð   Ô)ð+ð ! Ô*ð+ð ! Ô*ð+ð  ð+ð +ð +ð +ð +ð +ðZ˜œð ¨%¬,ð ð ð ð ð ð ð ð rT   rš   Ú	stem_typerU   c                 ó:   ‡ — t          ˆ fd„dD ¦   «         ¦  «        S )z®Check if stem type is deep (has multiple convolutions).

    Args:
        stem_type: Type of stem to check.

    Returns:
        True if stem is deep, False otherwise.
    c                 ó   •— g | ]}|‰v ‘ŒS © r±   )Ú.0Úsr®   s     €rS   ú
<listcomp>z is_stem_deep.<locals>.<listcomp>×  s   ø€ Ð;Ð;Ð; 1��Y�Ð;Ð;Ð;rT   )ÚdeepÚtiered)Úany)r®   s   `rS   Úis_stem_deepr¸   Î  s)   ø€ õ Ð;Ð;Ð;Ð;Ð(:Ð;Ñ;Ô;Ñ<Ô<Ð<rT   é@   Ú Tr=   r>   r-   r.   rA   r5   r6   c                 ó¼  — ||dœ}t          ¦   «         }	|dv sJ ‚t          |¦  «        r—d|v rd|z  dz  |dz  f}
n
|dz  |dz  f}
 || |
d         fdddœ|¤Ž|	d	<    ||
d         fi |¤Ž|	d
<    ||
d         |
d         fdddœ|¤Ž|	d<    ||
d         fi |¤Ž|	d<    ||
d         |fdddœ|¤Ž|	d<   |s ||fi |¤Ž|	d<   n || |fdddœ|¤Ž|	d<   |s ||fi |¤Ž|	d<   d|v r3t          j        dd¦  «        |	d<   t          j        ddd¬¦  «        |	d<   n5d|v rt          dddd¬¦  «        |	d<   nt          j        ddd¬¦  «        |	d<   t          j        |	¦  «        S )Nr:   )rº   ÚfixedÚsamerµ   Ú
deep_fixedÚ	deep_samer¶   r¶   rB   é   r‘   r   )Úkernel_sizer0   rH   rG   r   rJ   rI   ru   rt   é   r†   r‡   r¼   r,   Úpad)rÁ   r0   Úpaddingr•   r½   Úmax)r   r¸   rK   ÚConstantPad2dÚ	MaxPool2dr   r£   )r-   r.   r®   rA   r5   r6   r;   r<   rP   ÚstemÚstem_chss              rS   Úcreate_resnetv2_stemrÊ   Ú  s/  € ð  UÐ	+Ð	+€BÝ‰=Œ=€DØÐZÐZÐZÐZÐZõ �IÑÔð 5à�yÐ Ð Ø˜G™ qÑ(¨'°Q©,Ð7ˆHˆHà 1™ g°¡lÐ3ˆHØ"˜
 6¨8°A¬;ÐVÀAÈaÐVÐVÐSUÐVÐVˆˆW‰Ø"˜
 8¨A¤;Ð5Ð5°"Ð5Ð5ˆˆW‰Ø"˜
 8¨A¤;°¸´Ð[ÈÐSTÐ[Ð[ÐXZÐ[Ð[ˆˆW‰Ø"˜
 8¨A¤;Ð5Ð5°"Ð5Ð5ˆˆW‰Ø"˜
 8¨A¤;°ÐWÀQÈqÐWÐWÐTVÐWÐWˆˆW‰Øð 	6Ø&˜J wÐ5Ð5°"Ð5Ð5ˆD�‰Møð "�z &¨'ÐQ¸qÈÐQÐQÈbÐQÐQˆˆV‰Øð 	5Ø%˜: gÐ4Ð4°Ð4Ð4ˆD�‰Là�)ÐÐåÔ& q¨"Ñ-Ô-ˆˆU‰Ý”|°¸!ÀQÐGÑGÔGˆˆV‰ˆØ	�9Ð	Ð	å$ U¸À!ÈVÐTÑTÔTˆˆV‰ˆõ ”|°¸!ÀQÐGÑGÔGˆˆV‰åŒ=˜ÑÔÐrT   c            '       óš  ‡ — e Zd ZdZdddddddd	d
dd
dej         eed¬¦  «        eddd
ddfde	e
         dee
df         de
de
dede
de
de
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$„Zej        j        d@d#ed%dfd&„¦   «         Zej                             ¦   «         dAd(ed)ed%dfd*„¦   «         Zej        j        dBd+ed%eeef         fd,„¦   «         Zej        j        d@d-ed%dfd.„¦   «         Zej        j        d%ej        fd/„¦   «         ZdCde
dee         d%dfd0„Z	 	 	 	 	 dDd2ej        d3eee
e	e
         f                  d4ed5ed6ed7ed%ee	ej                 eej        e	ej                 f         f         fd8„Z 	 	 	 dEd3ee
e	e
         f         d9ed:efd;„Z!d2ej        d%ej        fd<„Z"dBd2ej        d=ed%ej        fd>„Z#d2ej        d%ej        fd?„Z$ˆ xZ%S )Fr(   z7Implementation of Pre-activation (v2) ResNet mode.
    )é   é   i   i   éè  rB   Úavgr=   r   r¹   rº   FTrr   r>   r,   NÚlayersÚchannels.Únum_classesÚin_chansÚglobal_poolÚoutput_strideÚwidth_factorrÉ   r®   rœ   rA   Úbasicr/   r4   r6   r5   Ú	drop_rater8   r[   c                 ó”  •— t          ¦   «                              ¦   «          ||dœ}|| _        || _        || _        |}t          ||¬¦  «        }t          |¦  «        }g | _        t          ||z  ¦  «        }t          |||	|f||dœ|¤Ž| _
        |rt          |	¦  «        rdndnd}| j                             t          |d|¬¦  «        ¦  «         |}d	}d
}t          ||d¬¦  «        }|r|rt          nt           }n|rJ ‚t"          }t%          j        ¦   «         | _        t+          t-          |||¦  «        ¦  «        D ]ž\  }\  }} }!t          | |z  ¦  «        }"|dk    rd
nd}#||k    r||#z  }d
}#t/          ||"f|#||||
||||!|dœ
|¤Ž}$|"}||#z  }| xj        t          ||d|› �¬¦  «        gz  c_        | j                             t3          |¦  «        |$¦  «         ŒŸ|x| _        | _        |r || j        fi |¤Žnt%          j        ¦   «         | _        t=          | j        |f|| j        ddœ|¤Ž| _        |                       |¬¦  «         dS )aÒ  
        Args:
            layers (List[int]) : number of layers in each block
            channels (List[int]) : number of channels in each block:
            num_classes (int): number of classification classes (default 1000)
            in_chans (int): number of input (color) channels. (default 3)
            global_pool (str): Global pooling type. One of 'avg', 'max', 'avgmax', 'catavgmax' (default 'avg')
            output_stride (int): output stride of the network, 32, 16, or 8. (default 32)
            width_factor (int): channel (width) multiplication factor
            stem_chs (int): stem width (default: 64)
            stem_type (str): stem type (default: '' == 7x7)
            avg_down (bool): average pooling in residual downsampling (default: False)
            preact (bool): pre-activation (default: True)
            act_layer (Union[str, nn.Module]): activation layer
            norm_layer (Union[str, nn.Module]): normalization layer
            conv_layer (nn.Module): convolution module
            drop_rate: classifier dropout rate (default: 0.)
            drop_path_rate: stochastic depth rate (default: 0.)
            zero_init_last: zero-init last weight in residual path (default: False)
        r:   )r4   )r5   r6   z
stem.conv3ú	stem.convz	stem.normr‘   )Únum_chsÚ	reductionÚmoduleé   r   T)Ú	stagewiser   )
r0   r1   r›   r/   rœ   r4   r5   r6   r�   rž   zstages.)Ú	pool_typerØ   Úuse_conv©r[   N)!rD   rE   rÒ   rÓ   rØ   r   r   Úfeature_infor   rÊ   rÈ   r¸   Úappendr¢   r   r*   rq   rz   rK   r£   ÚstagesÚ	enumerateÚziprš   r¦   r§   Únum_featuresÚhead_hidden_sizerL   r‡   r   ÚheadÚinit_weights)&rO   rÐ   rÑ   rÒ   rÓ   rÔ   rÕ   rÖ   rÉ   r®   rœ   rA   r×   r/   r4   r6   r5   rØ   r8   r[   r;   r<   rP   ÚwfÚ	stem_featr©   Úcurr_strider1   Ú
block_dprsrž   Ú	stage_idxÚdÚcÚbdprr.   r0   ÚstagerR   s&                                        €rS   rE   zResNetV2.__init__  s  ø€ õX 	‰Œ×ÒÑÔÐØ¨Ð/Ð/ˆØ&ˆÔØ ˆŒØ"ˆŒØˆÝ'¨
¸iÐHÑHÔHˆ
Ý! )Ñ,Ô,ˆ	àˆÔÝ! (¨R¡-Ñ0Ô0ˆÝ(ØØØØð	
ð
 "Ø!ð
ð 
ð ð
ð 
ˆŒ	ð SYÐi¥\°)Ñ%<Ô%<ÐM�\�\À+À+Ð^iˆ	ØÔ× Ò ¥¨hÀ!ÈIÐ!VÑ!VÔ!VÑWÔWÐWàˆØˆØˆÝ.¨~¸vÐQUÐVÑVÔVˆ
Øð 	"Ø&+ÐA•{�{Õ1AˆHˆHàÐÐ�9Ý!ˆHÝ”m‘o”oˆŒÝ'0µ°V¸XÀzÑ1RÔ1RÑ'SÔ'Sð 	:ð 	:Ñ#ˆI‘|˜˜1˜dÝ$ Q¨¡VÑ,Ô,ˆGØ# qš.˜.�Q�Q¨aˆFØ˜mÒ+Ð+Ø˜FÑ"�Ø�ÝØØðð Ø!ØØ)Ø!Ø#Ø%Ø%ØØ!ðð ð ðð ˆEð ˆHØ˜6Ñ!ˆKØÐÔ¥$¨xÀ;ÐWlÐajÐWlÐWlÐ"mÑ"mÔ"mÐ!nÑnÐÔØŒK×"Ò"¥3 y¡>¤>°5Ñ9Ô9Ð9Ð9à4<Ð<ˆÔ˜DÔ1Ø;AÐT�J�J˜tÔ0Ð7Ð7°BÐ7Ð7Ð7ÅrÄ{Á}Ä}ˆŒ	Ý"ØÔØð
ð "Ø”nØð
ð 
ð ð
ð 
ˆŒ	ð 	×Ò¨ÐÑ8Ô8Ð8Ð8Ð8rT   rU   c                 óN   — t          t          t          |¬¦  «        | ¦  «         dS )zInitialize model weights.râ   N)r#   r   Ú_init_weights)rO   r[   s     rS   rë   zResNetV2.init_weights‚  s'   € õ 	•G�M¸.ÐIÑIÔIÈ4ÑPÔPÐPÐPÐPrT   úresnet/Úcheckpoint_pathÚprefixc                 ó(   — t          | ||¦  «         dS )zâLoad model-specific weights from an original NumPy checkpoint.

        This helper handles the foreign ``.npz`` checkpoint layout. Native
        PyTorch state dictionaries use the generic factory/checkpoint loaders.
        N)Ú_load_weights)rO   rø   rù   s      rS   Úload_pretrainedzResNetV2.load_pretrained‡  s   € õ 	�d˜O¨VÑ4Ô4Ð4Ð4Ð4rT   Úcoarsec                 ó4   — t          d|rdnddg¬¦  «        }|S )z"Group parameters for optimization.z^stemz^stages\.(\d+))z^stages\.(\d+)\.blocks\.(\d+)N)z^norm)iŸ† )rÈ   r¤   )r¢   )rO   rý   Úmatchers      rS   Úgroup_matcherzResNetV2.group_matcher�  s;   € õ ØØ(.ð Ð$Ð$Ø8Ø$ð5ð
ñ 
ô 
ˆð ˆrT   Úenablec                 ó(   — | j         D ]	}||_        Œ
dS )z)Enable or disable gradient checkpointing.N)rå   r¡   )rO   r  r³   s      rS   Úset_grad_checkpointingzResNetV2.set_grad_checkpointingœ  s*   € ð ”ð 	*ð 	*ˆAØ#)ˆAÔ Ð ð	*ð 	*rT   c                 ó   — | j         j        S )zGet the classifier head.)rê   ÚfcrZ   s    rS   Úget_classifierzResNetV2.get_classifier¢  s   € ð ŒyŒ|ÐrT   c                 óZ   — t          | ¦  «        }|| _         | j        j        ||fi |¤Ž dS )zŸReset the classifier head.

        Args:
            num_classes: Number of classes for new classifier.
            global_pool: Global pooling type.
        N)r   rÒ   rê   Úreset)rO   rÒ   rÔ   rP   s       rS   Úreset_classifierzResNetV2.reset_classifier§  s=   € õ ˜dÑ#Ô#ˆØ&ˆÔØˆŒ	Œ˜ [Ð7Ð7°BÐ7Ð7Ð7Ð7Ð7rT   ÚNCHWr]   Úindicesr‡   Ú
stop_earlyÚ
output_fmtÚintermediates_onlyc                 óÂ  — |dv s
J d¦   «         ‚g }t          d|¦  «        \  }}	d}
|j        dd…         \  }}| j        D ]*} ||¦  «        }|j        dd…         |dz  |dz  fk    r|}Œ+|
|v r|                     |¦  «         t	          | j        ¦  «        }t          j                             ¦   «         s|s| j        }n| j        d|	…         }t          |d¬	¦  «        D ]^\  }
} ||¦  «        }|
|v rJ|
|k    r/|r|  
                    |¦  «        n|}|                     |¦  «         ŒI|                     |¦  «         Œ_|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.é   r   éþÿÿÿNr‘   r   )Ústart)r    ÚshaperÈ   rä   Úlenrå   rl   r¬   r­   ræ   r‡   )rO   r]   r  r‡   r  r  r  ÚintermediatesÚtake_indicesÚ	max_indexÚfeat_idxÚHÚWrÈ   Úx_downÚlast_idxrå   rô   Úx_inters                      rS   Úforward_intermediateszResNetV2.forward_intermediates²  s´  € ð* ˜YÐ&Ð&Ð&Ð(DÑ&Ô&Ð&ØˆÝ"6°q¸'Ñ"BÔ"BÑˆ�ið ˆØŒw�r�s�sŒ|‰ˆˆ1Ø”Ið 	ð 	ˆDØ��Q‘”ˆAØŒw�r�s�sŒ|  A¡ q¨1¡u˜~Ò-Ð-Ø�øØ�|Ð#Ð#Ø× Ò  Ñ(Ô(Ð(Ý�t”{Ñ#Ô#ˆÝŒ9×!Ò!Ñ#Ô#ð 	-¨:ð 	-Ø”[ˆFˆFà”[  ) Ô,ˆFå(¨°qÐ9Ñ9Ô9ð 	,ð 	,‰OˆH�eØ��a‘”ˆAØ˜<Ð'Ð'Ø˜xÒ'Ð'Ø.2Ð9˜dŸiši¨™lœl˜l¸�GØ!×(Ò(¨Ñ1Ô1Ð1Ð1à!×(Ò(¨Ñ+Ô+Ð+øàð 	!Ø Ð à�xÒÐØ—	’	˜!‘”ˆAà�-ÐÐrT   Ú
prune_normÚ
prune_headc                 ó¸   — t          d|¦  «        \  }}| j        d|…         | _        |rt          j        ¦   «         | _        |r|                      dd¦  «         |S )z@ Prune layers not required for specified intermediates.
        r  Nr   rº   )r    rå   rK   rL   r‡   r	  )rO   r  r  r   r  r  s         rS   Úprune_intermediate_layersz"ResNetV2.prune_intermediate_layersë  sc   € õ #7°q¸'Ñ"BÔ"BÑˆ�iØ”k * 9 *Ô-ˆŒØð 	&Ýœ™œˆDŒIØð 	)Ø×!Ò! ! RÑ(Ô(Ð(ØÐrT   c                 ó„   — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|S )z’Forward pass through feature extraction layers.

        Args:
            x: Input tensor.

        Returns:
            Feature tensor.
        )rÈ   rå   r‡   rŠ   s     rS   Úforward_featureszResNetV2.forward_featuresû  s6   € ð �IŠI�a‰LŒLˆØ�KŠK˜‰NŒNˆØ�IŠI�a‰LŒLˆØˆrT   Ú
pre_logitsc                 ó^   — |r|                       ||¬¦  «        n|                       |¦  «        S )zàForward pass through classifier head.

        Args:
            x: Input features.
            pre_logits: Return features before final linear layer.

        Returns:
            Classification logits or features.
        )r%  )rê   )rO   r]   r%  s      rS   Úforward_headzResNetV2.forward_head	  s0   € ð 7AÐRˆt�yŠy˜ zˆyÑ2Ô2Ð2ÀdÇiÂiÐPQÁlÄlÐRrT   c                 óZ   — |                       |¦  «        }|                      |¦  «        }|S )zoForward pass.

        Args:
            x: Input tensor.

        Returns:
            Output logits.
        )r$  r'  rŠ   s     rS   rc   zResNetV2.forward  s/   € ð ×!Ò! !Ñ$Ô$ˆØ×Ò˜aÑ Ô ˆØˆrT   )T©r÷   ©F)N)NFFr
  F)r   FT)&rf   rg   rh   ri   rK   r~   r   r   r   r   rj   r
   r§   r�   rk   r   rE   rl   r¬   Úignorerë   rü   r   r   r   r  ÚModuler  r	   r	  rm   r   r  r"  r$  r'  rc   rn   ro   s   @rS   r(   r(   
  s^  ø€ € € € € ðð ð )?Ø#ØØ$Ø!#Ø !ØØØ"ØØØ"&Ø"$¤'Ø#* 7¨<ÀBÐ#GÑ#GÔ#GØ#,Ø!Ø$&Ø#(ØØð-r9ð r9à˜”Iðr9ð ˜C ˜H”oðr9ð ð	r9ð
 ðr9ð ðr9ð ðr9ð ðr9ð ðr9ð ðr9ð ðr9ð ðr9ð ðr9ð  ðr9ð  ðr9ð  !ð!r9ð" !ð#r9ð$ ð%r9ð& "ð'r9ð( !ð)r9ð r9ð r9ð r9ð r9ð r9ðh „YÔðQð Q¨4ð Q¸4ð Qð Qð Qñ ÔðQð „Y×ÒÑÔð5ð 5¨sð 5¸Cð 5ÐPTð 5ð 5ð 5ñ Ôð5ð „YÔð	ð 	 Dð 	°T¸#¸s¸(´^ð 	ð 	ð 	ñ Ôð	ð „YÔð*ð *¨Tð *¸Tð *ð *ð *ñ Ôð*ð
 „YÔð ¤	ð ð ð ñ Ôðð	8ð 	8¨Cð 	8¸hÀs¼mð 	8ÐW[ð 	8ð 	8ð 	8ð 	8ð 8<ØØ$Ø$Ø',ð7 ð 7 àŒ|ð7 ð ˜e C¨¨c¬ NÔ3Ô4ð7 ð ð	7 ð
 ð7 ð ð7 ð !%ð7 ð 
ˆt�E”LÔ! 5¨¬°t¸E¼LÔ7IÐ)IÔ#JÐJÔ	Kð7 ð 7 ð 7 ð 7 ðv ./Ø$Ø#ð	ð à˜3  S¤	˜>Ô*ðð ðð ð	ð ð ð ð  %¤,ð °5´<ð ð ð ð ð
Sð 
S˜eœlð 
S¸ð 
SÈÌð 
Sð 
Sð 
Sð 
Sð˜œð ¨%¬,ð ð ð ð ð ð ð ð rT   rÝ   Únamer[   c                 ó:  — t          | t          j        ¦  «        sd|v rgt          | t          j        ¦  «        rMt          j                             | j        dd¬¦  «         t          j                             | j        ¦  «         dS t          | t          j        ¦  «        rVt          j         	                    | j        dd¬¦  «         | j        �&t          j                             | j        ¦  «         dS dS t          | t          j
        t          j        t          j        f¦  «        rJt          j                             | j        ¦  «         t          j                             | j        ¦  «         dS |r&t          | d	¦  «        r|                      ¦   «          dS dS dS )
z¯Initialize module weights.

    Args:
        module: PyTorch module to initialize.
        name: Module name.
        zero_init_last: Zero-initialize last layer weights.
    úhead.fcr,   g{®Gáz„?)ÚmeanÚstdÚfan_outÚrelu)ÚmodeÚnonlinearityNr[   )Ú
isinstancerK   ÚLinearÚConv2drW   Únormal_rY   rX   ÚbiasÚkaiming_normal_ÚBatchNorm2dÚ	LayerNormÚ	GroupNormÚones_Úhasattrr[   )rÝ   r-  r[   s      rS   rö   rö   #  sc  € õ �&�"œ)Ñ$Ô$ð  ¨°dÐ):Ð):½zÈ&ÕRTÔR[Ñ?\Ô?\Ð):Ý
Œ�Š˜œ¨C°TˆÑ:Ô:Ð:Ý
Œ�Š�v”{Ñ#Ô#Ð#Ð#Ð#Ý	�F�BœIÑ	&Ô	&ð  Ý
Œ×Ò ¤°IÈFÐÑSÔSÐSØŒ;Ð"ÝŒG�NŠN˜6œ;Ñ'Ô'Ð'Ð'Ð'ð #Ð"å	�F�Rœ^­R¬\½2¼<ÐHÑ	IÔ	Ið  Ý
Œ�Š�f”mÑ$Ô$Ð$Ý
Œ�Š�v”{Ñ#Ô#Ð#Ð#Ð#Ø	ð  �G FÐ,<Ñ=Ô=ð  Ø×ÒÑÔÐÐÐð ð  ð  ð  rT   r÷   Úmodelrø   rù   c                 óv  — d„ }t          ||› d�ft          | ¦  «        j        › d�¬¦  «        }t          | j        j        j        j        d          |||› d�         ¦  «        ¦  «        }| j        j        j                             |¦  «         | j	        j                              |||› d�         ¦  «        ¦  «         | j	        j
                              |||› d�         ¦  «        ¦  «         t          t          | j        dd ¦  «        t          j        ¦  «        r | j        j        j        j        d	         ||› d
�         j        d         k    rl| j        j        j                              |||› d
�         ¦  «        ¦  «         | j        j        j
                              |||› d�         ¦  «        ¦  «         t#          | j                             ¦   «         ¦  «        D �]R\  }\  }}t#          |j                             ¦   «         ¦  «        D �]!\  }	\  }
}d}|› d|dz   › d|	dz   d›d�}|j        j                              |||› d|› d�         ¦  «        ¦  «         |j        j                              |||› d|› d�         ¦  «        ¦  «         |j        j                              |||› d|› d�         ¦  «        ¦  «         |j        j                              |||› d�         ¦  «        ¦  «         |j        j                              |||› d�         ¦  «        ¦  «         |j        j                              |||› d�         ¦  «        ¦  «         |j        j
                              |||› d�         ¦  «        ¦  «         |j        j
                              |||› d�         ¦  «        ¦  «         |j        j
                              |||› d�         ¦  «        ¦  «         |j        �;||› d|› d�         }|j        j        j                              ||¦  «        ¦  «         �Œ#�ŒTd S )Nc                 ón   — | j         dk    r|                      g d¢¦  «        } t          j        | ¦  «        S )zPossibly convert HWIO to OIHW.rÞ   )rB   r‘   r   r   )ÚndimÚ	transposerl   Ú
from_numpy)Úconv_weightss    rS   Út2pz_load_weights.<locals>.t2p;  s8   € àÔ Ò!Ð!Ø'×1Ò1°,°,°,Ñ?Ô?ˆLÝÔ Ñ-Ô-Ð-rT   z%root_block/standardized_conv2d/kernelz.load_pretrained())Úrequired_anyÚcontextr   zgroup_norm/gammazgroup_norm/betar  r   zhead/conv2d/kerneléÿÿÿÿzhead/conv2d/biasÚstandardized_conv2dÚblockz/unitÚ02dú/za/z/kernelzb/zc/za/group_norm/gammazb/group_norm/gammazc/group_norm/gammaza/group_norm/betazb/group_norm/betazc/group_norm/betaza/proj/)r!   Útyperf   r$   rÈ   r†   rY   r  Úcopy_r‡   r:  r6  r�   rê   rK   r8  r  ræ   rå   Únamed_childrenr¤   rH   rJ   ru   rG   rI   rt   rF   )rA  rø   rù   rH  ÚweightsÚstem_conv_wÚiÚsnamerô   ÚjÚbnamerM  ÚcnameÚblock_prefixÚws                  rS   rû   rû   9  sw  € ð.ð .ð .õ #ØØÐFÐFÐFÐHÝ˜‘;”;Ô'Ð;Ð;Ð;ðñ ô €Gõ
 #ØŒ
ŒÔÔ$ QÔ'¨¨¨W¸Ð5eÐ5eÐ5eÔ-fÑ)gÔ)gñiô i€Kà	„J„OÔ× Ò  Ñ-Ô-Ð-Ø	„JÔ×Ò˜C˜C ¨6Ð(CÐ(CÐ(CÔ DÑEÔEÑFÔFÐFØ	„J„O×Ò˜#˜#˜g¨Ð&@Ð&@Ð&@ÔAÑBÔBÑCÔCÐCÝ•'˜%œ* d¨DÑ1Ô1µ2´9Ñ=Ô=ð LØŒJŒMÔ Ô& qÔ)¨W¸Ð5RÐ5RÐ5RÔ-SÔ-YÐZ\Ô-]Ò]Ð]ØŒ
ŒÔ×"Ò" 3 3 w°&Ð/LÐ/LÐ/LÔ'MÑ#NÔ#NÑOÔOÐOØŒ
ŒÔ× Ò    W°Ð-HÐ-HÐ-HÔ%IÑ!JÔ!JÑKÔKÐKÝ& u¤|×'BÒ'BÑ'DÔ'DÑEÔEð ;ñ ;Ñˆ‰>ˆE�5Ý!*¨5¬<×+FÒ+FÑ+HÔ+HÑ!IÔ!Ið 	;ñ 	;ÑˆA‰~��uØ)ˆEØ$ÐCÐC¨1¨q©5ÐCÐC°q¸1±uÐCÐCÐCÐCˆLØŒKÔ×$Ò$ S S¨°LÐ1RÐ1RÀEÐ1RÐ1RÐ1RÔ)SÑ%TÔ%TÑUÔUÐUØŒKÔ×$Ò$ S S¨°LÐ1RÐ1RÀEÐ1RÐ1RÐ1RÔ)SÑ%TÔ%TÑUÔUÐUØŒKÔ×$Ò$ S S¨°LÐ1RÐ1RÀEÐ1RÐ1RÐ1RÔ)SÑ%TÔ%TÑUÔUÐUØŒKÔ×$Ò$ S S¨°LÐ1TÐ1TÐ1TÔ)UÑ%VÔ%VÑWÔWÐWØŒKÔ×$Ò$ S S¨°LÐ1TÐ1TÐ1TÔ)UÑ%VÔ%VÑWÔWÐWØŒKÔ×$Ò$ S S¨°LÐ1TÐ1TÐ1TÔ)UÑ%VÔ%VÑWÔWÐWØŒKÔ×"Ò" 3 3 w°,Ð/QÐ/QÐ/QÔ'RÑ#SÔ#SÑTÔTÐTØŒKÔ×"Ò" 3 3 w°,Ð/QÐ/QÐ/QÔ'RÑ#SÔ#SÑTÔTÐTØŒKÔ×"Ò" 3 3 w°,Ð/QÐ/QÐ/QÔ'RÑ#SÔ#SÑTÔTÐTØÔÐ+Ø˜|ÐBÐB°EÐBÐBÐBÔC�ØÔ Ô%Ô,×2Ò2°3°3°q±6´6Ñ:Ô:Ð:ùñ	;ð;ð ;rT   FÚvariantÚ
pretrainedÚkwargsc                 óN   — t          d¬¦  «        }t          t          | |fd|i|¤ŽS )z×Create a ResNetV2 model.

    Args:
        variant: Model variant name.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        ResNetV2 model instance.
    T)Úflatten_sequentialÚfeature_cfg)r¢   r   r(   )r\  r]  r^  ra  s       rS   Ú_create_resnetv2rb  a  sD   € õ ¨$Ð/Ñ/Ô/€KÝÝ�'˜:ðð àðð ðð ð rT   c           	      óL   — t          | f|dt          t          d¬¦  «        dœ|¤ŽS )zèCreate a ResNetV2 model with BiT weights.

    Args:
        variant: Model variant name.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        ResNetV2 model instance.
    r¼   g:Œ0âŽyE>)Úeps)r]  r®   r5   )rb  r   r   )r\  r]  r^  s      rS   Ú_create_resnetv2_bitre  t  sC   € õ ØðàØÝ�9¨$Ð/Ñ/Ô/ð	ð ð
 ðð ð rT   Úurlc                 ó6   — | dddddt           t          dddd	œ|¥S )
NrÎ   )rB   éà   rh  )rÂ   rÂ   g      ì?ÚbilinearrÚ   r/  z
apache-2.0)rf  rÒ   Ú
input_sizeÚ	pool_sizeÚcrop_pctÚinterpolationr0  r1  Ú
first_convÚ
classifierÚlicenser   )rf  r^  s     rS   Ú_cfgrq  ˆ  s8   € àØ¨=ÀvØ¨JÝ'Õ0FØ!°Øðð ð ðð rT   z%resnetv2_50x1_bit.goog_distilled_in1kztimm/Úbicubic)Ú	hf_hub_idrm  Úcustom_loadz-resnetv2_152x2_bit.goog_teacher_in21k_ft_in1kz1resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k_384)rB   é€  ru  )é   rv  r+   )rs  rj  rk  rl  rm  rt  z$resnetv2_50x1_bit.goog_in21k_ft_in1k)rB   éÀ  rw  )é   rx  )rs  rj  rk  rl  rt  z$resnetv2_50x3_bit.goog_in21k_ft_in1kz%resnetv2_101x1_bit.goog_in21k_ft_in1kz%resnetv2_101x3_bit.goog_in21k_ft_in1kz%resnetv2_152x2_bit.goog_in21k_ft_in1kz%resnetv2_152x4_bit.goog_in21k_ft_in1k)rB   éà  ry  )é   rz  zresnetv2_50x1_bit.goog_in21kiSU  )rs  rÒ   rt  zresnetv2_50x3_bit.goog_in21kzresnetv2_101x1_bit.goog_in21kzresnetv2_101x3_bit.goog_in21kzresnetv2_152x2_bit.goog_in21kzresnetv2_152x4_bit.goog_in21kzresnetv2_18.ra4_e3600_r224_in1kgÍÌÌÌÌÌì?)rB   é   r{  )rs  rm  rl  Útest_input_sizeÚtest_crop_pctz resnetv2_18d.ra4_e3600_r224_in1kz
stem.conv1)rs  rm  rl  r|  r}  rn  )rs  rl  rj  rk  r|  rm  rn  gffffffî?)rm  rn  )rm  )rs  rm  rn  rl  r|  r}  )zresnetv2_34.ra4_e3600_r224_in1kz resnetv2_34d.ra4_e3600_r224_in1kz resnetv2_34d.ra4_e3600_r384_in1kzresnetv2_50.a1h_in1kzresnetv2_50d.untrainedzresnetv2_50t.untrainedzresnetv2_101.a1h_in1kzresnetv2_101d.untrainedzresnetv2_152.untrainedzresnetv2_152d.untrainedzresnetv2_50d_gn.ah_in1kzresnetv2_50d_evos.ah_in1kzresnetv2_50d_frn.untrainedc                 ó(   — t          	 d| g d¢ddœ|¤ŽS )zResNetV2-50x1-BiT model.Úresnetv2_50x1_bit©rB   rÞ   é   rB   r   ©r]  rÐ   rÖ   )r  ©re  ©r]  r^  s     rS   r  r  ø  ó=   € õ  ØðcØ(2¸<¸<¸<ÐVWðcð cØ[aðcð cð crT   c                 ó(   — t          	 d| g d¢ddœ|¤ŽS )zResNetV2-50x3-BiT model.Úresnetv2_50x3_bitr€  rB   r‚  )r‡  rƒ  r„  s     rS   r‡  r‡  ÿ  r…  rT   c                 ó(   — t          	 d| g d¢ddœ|¤ŽS )zResNetV2-101x1-BiT model.Úresnetv2_101x1_bit©rB   rÞ   é   rB   r   r‚  )r‰  rƒ  r„  s     rS   r‰  r‰    ó=   € õ  ØðeØ)3¸M¸M¸MÐXYðeð eØ]cðeð eð erT   c                 ó(   — t          	 d| g d¢ddœ|¤ŽS )zResNetV2-101x3-BiT model.Úresnetv2_101x3_bitrŠ  rB   r‚  )rŽ  rƒ  r„  s     rS   rŽ  rŽ    rŒ  rT   c                 ó(   — t          	 d| g d¢ddœ|¤ŽS )zResNetV2-152x2-BiT model.Úresnetv2_152x2_bit©rB   rÀ   é$   rB   r‘   r‚  )r�  rƒ  r„  s     rS   r�  r�    rŒ  rT   c                 ó(   — t          	 d| g d¢ddœ|¤ŽS )zResNetV2-152x4-BiT model.Úresnetv2_152x4_bitr‘  rÞ   r‚  )r”  rƒ  r„  s     rS   r”  r”    rŒ  rT   c           	      óv   — t          g d¢dddt          t          ¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )	zResNetV2-18 model.©r‘   r‘   r‘   r‘   ©r¹   é€   rÌ   rÍ   Tr+   ©rÐ   rÑ   r×   r/   r5   r6   Úresnetv2_18r]  )rš  ©r¢   r   r   rb  ©r]  r^  Ú
model_argss      rS   rš  rš  "  sX   € õ Øˆ|ˆ|Ð&9ÀÐTWÝ ­^ðñ ô €Jõ Ð_Ð_°jÐ_ÅDÈÐD^ÐD^ÐW]ÐD^ÐD^Ð_Ð_Ð_rT   c           
      óz   — t          g d¢dddt          t          dd¬¦  «        }t          d	d| it          |fi |¤Ž¤ŽS )
z'ResNetV2-18d model (deep stem variant).r–  r—  Tr+   rµ   ©rÐ   rÑ   r×   r/   r5   r6   r®   rœ   Úresnetv2_18dr]  )r   r›  rœ  s      rS   r   r   ,  s]   € õ Øˆ|ˆ|Ð&9ÀÐTWÝ ­^ÀvÐX\ðñ ô €Jõ Ð`Ð`°zÐ`ÅTÈ*ÐE_ÐE_ÐX^ÐE_ÐE_Ð`Ð`Ð`rT   c           	      ór   — t          ddddt          t          ¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )	zResNetV2-34 model.r€  r—  Tr+   r™  Úresnetv2_34r]  )r¢  r›  rœ  s      rS   r¢  r¢  6  sT   € õ ØÐ&9ÀÐTWÝ ­^ðñ ô €Jõ Ð_Ð_°jÐ_ÅDÈÐD^ÐD^ÐW]ÐD^ÐD^Ð_Ð_Ð_rT   c           
      óv   — t          ddddt          t          dd¬¦  «        }t          d	d| it          |fi |¤Ž¤ŽS )
z'ResNetV2-34d model (deep stem variant).r€  r—  Tr+   rµ   rŸ  Úresnetv2_34dr]  )r¤  r›  rœ  s      rS   r¤  r¤  @  sY   € õ ØÐ&9ÀÐTWÝ ­^ÀvÐX\ðñ ô €Jõ Ð`Ð`°zÐ`ÅTÈ*ÐE_ÐE_ÐX^ÐE_ÐE_Ð`Ð`Ð`rT   c           	      óp   — t          g d¢t          t          ¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )zResNetV2-50 model.r€  ©rÐ   r5   r6   Úresnetv2_50r]  )r§  r›  rœ  s      rS   r§  r§  J  sG   € õ ˜\˜\˜\µmÕP^Ð_Ñ_Ô_€JÝÐ_Ð_°jÐ_ÅDÈÐD^ÐD^ÐW]ÐD^ÐD^Ð_Ð_Ð_rT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z'ResNetV2-50d model (deep stem variant).r€  rµ   T©rÐ   r5   r6   r®   rœ   Úresnetv2_50dr]  )rª  r›  rœ  s      rS   rª  rª  Q  sT   € õ Øˆ|ˆ|­Å.Ø 4ð)ñ )ô )€Jõ Ð`Ð`°zÐ`ÅTÈ*ÐE_ÐE_ÐX^ÐE_ÐE_Ð`Ð`Ð`rT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z)ResNetV2-50t model (tiered stem variant).r€  r¶   Tr©  Úresnetv2_50tr]  )r¬  r›  rœ  s      rS   r¬  r¬  Z  sT   € õ Øˆ|ˆ|­Å.Ø Tð+ñ +ô +€Jõ Ð`Ð`°zÐ`ÅTÈ*ÐE_ÐE_ÐX^ÐE_ÐE_Ð`Ð`Ð`rT   c           	      óp   — t          g d¢t          t          ¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )zResNetV2-101 model.rŠ  r¦  Úresnetv2_101r]  )r®  r›  rœ  s      rS   r®  r®  c  óG   € õ ˜]˜]˜]µ}ÕQ_Ð`Ñ`Ô`€JÝÐ`Ð`°zÐ`ÅTÈ*ÐE_ÐE_ÐX^ÐE_ÐE_Ð`Ð`Ð`rT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z(ResNetV2-101d model (deep stem variant).rŠ  rµ   Tr©  Úresnetv2_101dr]  )r±  r›  rœ  s      rS   r±  r±  j  óT   € õ Øˆ}ˆ}­Å>Ø 4ð)ñ )ô )€Jõ ÐaÐa¸
ÐaÅdÈ:ÐF`ÐF`ÐY_ÐF`ÐF`ÐaÐaÐarT   c           	      óp   — t          g d¢t          t          ¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )zResNetV2-152 model.r‘  r¦  Úresnetv2_152r]  )r´  r›  rœ  s      rS   r´  r´  s  r¯  rT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z(ResNetV2-152d model (deep stem variant).r‘  rµ   Tr©  Úresnetv2_152dr]  )r¶  r›  rœ  s      rS   r¶  r¶  z  r²  rT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z,ResNetV2-50d model with Group Normalization.r€  rµ   Tr©  Úresnetv2_50d_gnr]  )r¸  )r¢   r   r   rb  rœ  s      rS   r¸  r¸  …  sT   € õ Øˆ|ˆ|­Å,Ø 4ð)ñ )ô )€Jõ ÐcÐc¸*ÐcÍÈZÐHbÐHbÐ[aÐHbÐHbÐcÐcÐcrT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z ResNetV2-50d model with EvoNorm.r€  rµ   Tr©  Úresnetv2_50d_evosr]  )rº  )r¢   r   r   rb  rœ  s      rS   rº  rº  Ž  sT   € õ Øˆ|ˆ|­Å+Ø 4ð)ñ )ô )€Jõ ÐeÐe¸JÐeÍ$ÈzÐJdÐJdÐ]cÐJdÐJdÐeÐeÐerT   c           	      ót   — t          g d¢t          t          dd¬¦  «        }t          dd| it          |fi |¤Ž¤ŽS )z6ResNetV2-50d model with Filter Response Normalization.r€  rµ   Tr©  Úresnetv2_50d_frnr]  )r¼  )r¢   r   r   rb  rœ  s      rS   r¼  r¼  —  sU   € õ Øˆ|ˆ|­ÕBYØ 4ð)ñ )ô )€Jõ ÐdÐd¸:ÐdÍÈjÐIcÐIcÐ\bÐIcÐIcÐdÐdÐdrT   )Úresnetv2_50x1_bitmÚresnetv2_50x3_bitmÚresnetv2_101x1_bitmÚresnetv2_101x3_bitmÚresnetv2_152x2_bitmÚresnetv2_152x4_bitmÚresnetv2_50x1_bitm_in21kÚresnetv2_50x3_bitm_in21kÚresnetv2_101x1_bitm_in21kÚresnetv2_101x3_bitm_in21kÚresnetv2_152x2_bitm_in21kÚresnetv2_152x4_bitm_in21kÚresnetv2_50x1_bit_distilledÚresnetv2_152x2_bit_teacherÚresnetv2_152x2_bit_teacher_384)rº   Tr)  r*  )rº   )\ri   Úcollectionsr   Ú	functoolsr   Útypingr   r   r   r   r	   r
   r   rl   Útorch.nnrK   Ú	timm.datar   r   Útimm.layersr   r   r   r   r   r   r   r   r   r   r   r   r   r   r   Ú_builderr   Ú	_featuresr    Ú_helpersr!   Ú_manipulater"   r#   r$   Ú	_registryr%   r&   r'   Ú__all__r,  r*   rq   rz   r„   r�   rš   r§   r�   r¸   rj   r£   rÊ   r(   rö   Úno_gradrû   rb  re  rq  Údefault_cfgsr  r‡  r‰  rŽ  r�  r”  rš  r   r¢  r¤  r§  rª  r¬  r®  r±  r´  r¶  r¸  rº  r¼  rf   r±   rT   rS   ú<module>rÚ     s<  ððð ð> $Ð #Ð #Ð #Ð #Ð #Ø Ð Ð Ð Ð Ð Ø DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ DÐ Dà €€€Ø Ð Ð Ð Ð Ð à EÐ EÐ EÐ EÐ EÐ EÐ EÐ Eðeð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eð eà *Ð *Ð *Ð *Ð *Ð *Ø +Ð +Ð +Ð +Ð +Ð +Ø *Ð *Ð *Ð *Ð *Ð *Ø FÐ FÐ FÐ FÐ FÐ FÐ FÐ FÐ FÐ FØ YÐ YÐ YÐ YÐ YÐ YÐ YÐ YÐ YÐ Yàˆ,€ðYð Yð Yð Yð Y�"”)ñ Yô Yð Yðxbð bð bð bð b�r”yñ bô bð bðJPð Pð Pð Pð P�”ñ Pô Pð Pðf'ð 'ð 'ð 'ð '�R”Yñ 'ô 'ð 'ðB$2ð $2ð $2ð $2ð $2�B”Iñ $2ô $2ð $2ðN<ð <ð <ð <ð <�"”)ñ <ô <ð <ð~	=˜Cð 	= Dð 	=ð 	=ð 	=ð 	=ð ØØØ(Ø&˜w |ÀÐCÑCÔCØØð-ð -Øð-àð-ð ð-ð ð	-ð
 ð-ð ð-ð „]ð-ð -ð -ð -ð`Vð Vð Vð Vð VˆrŒyñ Vô Vð Vðr ð  ˜"œ)ð  ¨3ð  ÀTð  ÐUYð  ð  ð  ð  ð, €„�„ð$;ð $;˜œð $;°Sð $;À#ð $;ð $;ð $;ñ „ð$;ðNð ˜cð ¨tð Àsð Èxð ð ð ð ð&ð  #ð °4ð È3ð ÐS[ð ð ð ð ð(	ð 	ˆcð 	 #ð 	¨$¨s°C¨x¬.ð 	ð 	ð 	ð 	ð %Ð$ð a&à+¨T¨TØØ¨Tð.3ñ .3ô .3ða&ð
 4°T°TØØ¨Tð63ñ 63ô 63ða&ð 8¸¸ØØ ¨H¸sÐR[Ðimð:oñ :oô :oða&ð +¨D¨DØØ ¨H¸sÐPTð-Vñ -Vô -Vða&ð  +¨D¨DØØ ¨H¸sÐPTð-Vñ -Vô -Vð!a&ð& ,¨T¨TØØ ¨H¸sÐPTð.Vñ .Vô .Vð'a&ð, ,¨T¨TØØ ¨H¸sÐPTð.Vñ .Vô .Vð-a&ð2 ,¨T¨TØØ ¨H¸sÐPTð.Vñ .Vô .Vð3a&ð8 ,¨T¨TØØ ¨H¸sÐPTð.Vñ .Vô .Vð9a&ðB # D DØØ tð%-ñ %-ô %-ðCa&ðH # D DØØ tð%-ñ %-ô %-ðIa&ðN $ T TØØ tð&-ñ &-ô &-ðOa&ðT $ T TØØ tð&-ñ &-ô &-ðUa&ðZ $ T TØØ tð&-ñ &-ô &-ð[a&ð` $ T TØØ tð&-ñ &-ô &-ðaa&ðh & t tØØ¨#¸}Ð\_ð(añ (aô (aðia&ðn '¨¨ØØ¨#¸}Ð\_Øð)!ñ )!ô )!ðoa&ðv (, tØØ¨#¸}Ð\_ð(añ (aô (að )-¨ØØ¨#¸}Ð\_Øð)!ñ )!ô )!ð )-¨ØØ ¸(ÐTaØ¨Lð):ñ ):ô ):ð !˜DØØ¨$ÀÐ]`ðbñ bô bð #˜dØ¨Lð:ñ :ô :à"˜dØ¨Lð:ñ :ô :à!˜TØØ¨$ÀÐ]`ðbñ bô bð  $˜tØ¨Lð :ñ  :ô  :à"˜dØð!ñ !ô !à#˜tØ¨Lð :ñ  :ô  :ð  $˜tØØ¨LØ }ÀCð Iñ  Iô  Ið "& ØØ¨LØ }ÀCð"Iñ "Iô "Ið #' $Ø¨Lð#:ñ #:ô #:ða&ð a&ð a&ñ aô a€ðH ðcð c $ð c¸#ð cÀ(ð cð cð cñ „ðcð ðcð c $ð c¸#ð cÀ(ð cð cð cñ „ðcð ðeð e 4ð e¸3ð eÀ8ð eð eð eñ „ðeð ðeð e 4ð e¸3ð eÀ8ð eð eð eñ „ðeð ðeð e 4ð e¸3ð eÀ8ð eð eð eñ „ðeð ðeð e 4ð e¸3ð eÀ8ð eð eð eñ „ðeð ð`ð `˜Dð `°Cð `¸Hð `ð `ð `ñ „ð`ð ðað a˜Tð a°Sð a¸Xð að að añ „ðað ð`ð `˜Dð `°Cð `¸Hð `ð `ð `ñ „ð`ð ðað a˜Tð a°Sð a¸Xð að að añ „ðað ð`ð `˜Dð `°Cð `¸Hð `ð `ð `ñ „ð`ð ðað a˜Tð a°Sð a¸Xð að að añ „ðað ðað a˜Tð a°Sð a¸Xð að að añ „ðað ðað a˜Tð a°Sð a¸Xð að að añ „ðað ðbð b˜dð b°cð b¸hð bð bð bñ „ðbð ðað a˜Tð a°Sð a¸Xð að að añ „ðað ðbð b˜dð b°cð b¸hð bð bð bñ „ðbð ðdð d ð d¸ð dÀð dð dð dñ „ðdð ðfð f $ð f¸#ð fÀ(ð fð fð fñ „ðfð ðeð e ð e¸ð eÀð eð eð eñ „ðeð Ð ˜HØ@Ø@ØBØBØBØBØ >Ø >Ø!@Ø!@Ø!@Ø!@Ø#JØ"QØ&Yð'ð 'ñ ô ð ð ð rT   