§
    ·¯¸jò  ã                   óâ  — d 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c mZ ddlmZmZmZmZ ddlmZmZmZmZmZmZmZ ddlmZm Z  dd	l!m"Z" dd
l#m$Z$m%Z%m&Z&m'Z'm(Z(m)Z)m*Z*m+Z+ ddl,m-Z-m.Z.m/Z/ ddl0m1Z1m2Z2 ddl3m4Z4m5Z5m6Z6 ddgZ7 G d„ dej8        ¦  «        Z9 G d„ dej8        ¦  «        Z:d¶de;de<de9fd„Z=	 d·de;de>de<de9fd„Z?	 	 	 	 d¸de;de>de>dee@         de<de9fd„ZAd·de;de>de<de9fd„ZBd·de;de>de<de9fd„ZC	 	 	 d¹de;de>dee@         de<de9f
d„ZDdºd!e;dee;ef         fd"„ZE e4i d# eEd ¬$¦  «        “d% eEd&d'd(¬)¦  «        “d* eEd(d&eed+d,d¬-¦  «        “d. eEd/d0d1d2d3d4d(¬5¦  «        “d6 eEd7d(d2d3d/d0d1d8¬9¦  «        “d: eEd(eed,d+d;d<d¬=¦  «        “d> eEd?d(d&¬@¦  «        “dA eEdBd(d&¬@¦  «        “dC eEdDd(d&¬@¦  «        “dE eEdFd(d&¬@¦  «        “dG eEdHd(ee¬I¦  «        “dJ eEdKd(ee¬I¦  «        “dL eEdMd(ee¬I¦  «        “dN eEdOd(ee¬I¦  «        “dP eEdQd(ee¬I¦  «        “dR eEdSd(ee¬I¦  «        “dT eEdUd(d,d+¬V¦  «        “i dW eEdXd(d,d+¬V¦  «        “dY eEdZd(d[d\d+d;¬]¦  «        “d^ eE¦   «         “d_ eEd`d(d&¬@¦  «        “da eEdbd(d&¬@¦  «        “dc eEddd(d&¬@¦  «        “de eE¦   «         “df eEeed,d+d&¬g¦  «        “dh eEd(eed,d+d&¬i¦  «        “dj eEd(d,d+d&¬k¦  «        “dl eEd(d,d+d&¬k¦  «        “dm eEd(eed,d;d+d<dd&¬n¦	  «	        “do eEd(d,d;d+d<dd&¬p¦  «        “dq eEd(d+d,dd&¬r¦  «        “ds eEd(dtdud+d&¬v¦  «        “dw eEd(dxdtdudd&¬y¦  «        “dz eEd(dxdtdudd&¬y¦  «        “¥i d{ eEd(dxdtdudd&¬y¦  «        “d| eEd(dtdud+d}dd&¬p¦  «        “d~ eEd(d,d;d+d<dd&¬p¦  «        “d eEd(d,d;d+d<dd&¬p¦  «        “d€ eEd(d,d;d+d<dd&¬p¦  «        “d� eEd(dtdud+d}dd&¬p¦  «        “d‚ eEd(d+d,dd&¬r¦  «        “dƒ eEd(dxd,d;d+d<dd&¬„¦  «        “d… eEd(dtdud+d}dd&¬p¦  «        “d† eEd(dtdud+d}dd&¬p¦  «        “d‡ eEd,d;d+d&¬ˆ¦  «        “d‰ eEd(d+d,dd&¬r¦  «        “dŠ eEd(d}d‹d+dŒdd&¬p¦  «        “d� eEd(dtdud+dŽdd&¬p¦  «        “d� eEd(dtdud+dŽdd&¬p¦  «        “d� eEd(dxdtdud+d}dd&¬„¦  «        “d‘ eEd+d&¬’¦  «        “¥d“ eEd,d;d+d&¬ˆ¦  «        i¥¦  «        ZFe5d¶de<de9fd”„¦   «         ZGe5d¶de<de9fd•„¦   «         ZHe5d¶de<de9fd–„¦   «         ZIe5d¶de<de9fd—„¦   «         ZJe5d¶de<de9fd˜„¦   «         ZKe5d¶de<de9fd™„¦   «         ZLe5d¶de<de9fdš„¦   «         ZMe5d¶de<de9fd›„¦   «         ZNe5d¶de<de9fdœ„¦   «         ZOe5d¶de<de9fd�„¦   «         ZPe5d¶de<de9fdž„¦   «         ZQe5d¶de<de9fdŸ„¦   «         ZRe5d¶de<de9fd „¦   «         ZSe5d¶de<de9fd¡„¦   «         ZTe5d¶de<de9fd¢„¦   «         ZUe5d¶de<de9fd£„¦   «         ZVe5d¶de<de9fd¤„¦   «         ZWe5d¶de<de9fd¥„¦   «         ZXe5d¶de<de9fd¦„¦   «         ZYe5d¶de<de9fd§„¦   «         ZZe5d¶de<de9fd¨„¦   «         Z[e5d¶de<de9fd©„¦   «         Z\e5d¶de<de9fdª„¦   «         Z]e5d¶de<de9fd«„¦   «         Z^e5d¶de<de9fd¬„¦   «         Z_e5d¶de<de9fd­„¦   «         Z`e5d¶de<de9fd®„¦   «         Zae5d¶de<de9fd¯„¦   «         Zbe5d¶de<de9fd°„¦   «         Zce5d¶de<de9fd±„¦   «         Zde5d¶de<de9fd²„¦   «         Zee5d¶de<de9fd³„¦   «         Zfe5d¶de<de9fd´„¦   «         Zg e6ehd.d6dµœ¦  «         dS )»z× MobileNet V3

A PyTorch impl of MobileNet-V3, compatible with TF weights from official impl.

Paper: Searching for MobileNetV3 - https://arxiv.org/abs/1905.02244

Hacked together by / Copyright 2019, Ross Wightman
é    )Úpartial)ÚAnyÚDictÚCallableÚListÚOptionalÚTupleÚUnionN)ÚIMAGENET_DEFAULT_MEANÚIMAGENET_DEFAULT_STDÚIMAGENET_INCEPTION_MEANÚIMAGENET_INCEPTION_STD)ÚSelectAdaptivePool2dÚLinearÚ	LayerTypeÚPadTypeÚcreate_conv2dÚget_norm_act_layerÚget_device_dtypeé   )Úbuild_model_with_cfgÚpretrained_cfg_for_features)ÚSqueezeExcite)Ú	BlockArgsÚEfficientNetBuilderÚdecode_arch_defÚefficientnet_init_weightsÚround_channelsÚresolve_bn_argsÚresolve_act_layerÚBN_EPS_TF_DEFAULT)ÚFeatureInfoÚFeatureHooksÚfeature_take_indices)Úcheckpoint_seqÚ
checkpoint)Úgenerate_default_cfgsÚregister_modelÚregister_model_deprecationsÚMobileNetV3ÚMobileNetV3Featuresc            '       ó6  ‡ — e Zd ZdZddddddddd	d	d	d	ded
d
d	dd	d	fdededededededededede	e
         de	e
         de	e
         de	e
         dedededede	e         def&ˆ fd„Zd ej        fd!„Zej        j        d9d"ed eeef         fd#„¦   «         Zej        j        d:d$ed d	fd%„¦   «         Zej        j        d ej        fd&„¦   «         Zd;deded d	fd'„Z	 	 	 	 	 	 d<d)ej        d*e	eeee         f                  d+ed,ed-ed.ed/ed eeej                 eej        eej                 f         f         fd0„Z	 	 	 	 d=d*eeee         f         d2ed3ed/ed ee         f
d4„Z d)ej        d ej        fd5„Z!d9d)ej        d6ed ej        fd7„Z"d)ej        d ej        fd8„Z#ˆ xZ$S )>r*   a½  MobileNetV3.

    Based on my EfficientNet implementation and building blocks, this model utilizes the MobileNet-v3 specific
    'efficient head', where global pooling is done before the head convolution without a final batch-norm
    layer before the classifier.

    Paper: `Searching for MobileNetV3` - https://arxiv.org/abs/1905.02244

    Other architectures utilizing MobileNet-V3 efficient head that are supported by this impl include:
      * HardCoRe-NAS - https://arxiv.org/abs/2102.11646 (defn in hardcorenas.py uses this class)
      * FBNet-V3 - https://arxiv.org/abs/2006.02049
      * LCNet - https://arxiv.org/abs/2109.15099
      * MobileNet-V4 - https://arxiv.org/abs/2404.10518
    éè  é   é   Fé   TÚ Nç        ÚavgÚ
block_argsÚnum_classesÚin_chansÚ	stem_sizeÚfix_stemÚnum_featuresÚ	head_biasÚ	head_normÚpad_typeÚ	act_layerÚ
norm_layerÚaa_layerÚse_layerÚse_from_expÚround_chs_fnÚ	drop_rateÚdrop_path_rateÚlayer_scale_init_valueÚglobal_poolc                 ó”  •— t          ¦   «                              ¦   «          ||dœ}|
pt          j        }
|pt          j        }t          ||
¦  «        }|pt          }|| _        || _        || _	        d| _
        |s ||¦  «        }t          ||dfd|	dœ|¤Ž| _         ||fddi|¤Ž| _        t          dd|	|||
|||||d	œ
|¤Ž}t          j         |||¦  «        Ž | _        |j        | _        d
„ | j        D ¦   «         | _        |j        | _        || _        t/          |¬¦  «        | _        | j        | j                             ¦   «         z  }|rHt          || j        df|	ddœ|¤Ž| _         || j        fi |¤Ž| _        t          j        ¦   «         | _        nEt          || j        df|	|dœ|¤Ž| _        t          j        ¦   «         | _         |
d¬¦  «        | _        |rt          j        d¦  «        nt          j        ¦   «         | _        |dk    rtA          | j        |fi |¤Žnt          j        ¦   «         | _!        tE          | ¦  «         dS )a0  Initialize MobileNetV3.

        Args:
            block_args: Arguments for blocks of the network.
            num_classes: Number of classes for classification head.
            in_chans: Number of input image channels.
            stem_size: Number of output channels of the initial stem convolution.
            fix_stem: If True, don't scale stem by round_chs_fn.
            num_features: Number of output channels of the conv head layer.
            head_bias: If True, add a learnable bias to the conv head layer.
            head_norm: If True, add normalization to the head layer.
            pad_type: Type of padding to use for convolution layers.
            act_layer: Type of activation layer.
            norm_layer: Type of normalization layer.
            aa_layer: Type of anti-aliasing layer.
            se_layer: Type of Squeeze-and-Excite layer.
            se_from_exp: If True, calculate SE channel reduction from expanded mid channels.
            round_chs_fn: Callable to round number of filters based on depth multiplier.
            drop_rate: Dropout rate.
            drop_path_rate: Stochastic depth rate.
            layer_scale_init_value: Enable layer scale on compatible blocks if not None.
            global_pool: Type of pooling to use for global pooling features of the FC head.
        ©ÚdeviceÚdtypeFr.   é   ©ÚstrideÚpaddingÚinplaceTé    )
Úoutput_strider<   rB   rA   r=   r>   r?   r@   rD   rE   c                 ó   — g | ]
}|d          ‘ŒS )Ústage© ©Ú.0Úfs     úg/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/timm/models/mobilenetv3.pyú
<listcomp>z(MobileNetV3.__init__.<locals>.<listcomp>}   s   € ÐAÐAÐA¨!˜1˜Wœ:ÐAÐAÐAó    ©Ú	pool_typer   )rN   Úbias©rO   r   NrT   )#ÚsuperÚ__init__ÚnnÚReLUÚBatchNorm2dr   r   r5   r6   rC   Úgrad_checkpointingr   Ú	conv_stemÚbn1r   Ú
SequentialÚblocksÚfeaturesÚfeature_infoÚ
stage_endsÚin_chsr9   Úhead_hidden_sizer   rF   Ú	feat_multÚ	conv_headÚ	norm_headÚIdentityÚact2ÚFlattenÚflattenr   Ú
classifierr   )Úselfr4   r5   r6   r7   r8   r9   r:   r;   r<   r=   r>   r?   r@   rA   rB   rC   rD   rE   rF   rI   rJ   ÚddÚnorm_act_layerÚbuilderÚnum_pooled_chsÚ	__class__s                             €rX   r`   zMobileNetV3.__init__-   sÊ  ø€ õ^ 	‰Œ×ÒÑÔÐØ¨Ð/Ð/ˆØÐ(¥¤ˆ	ØÐ1¥2¤>ˆ
Ý+¨J¸	ÑBÔBˆØÐ,�}ˆØ&ˆÔØ ˆŒØ"ˆŒØ"'ˆÔð ð 	0Ø$˜ YÑ/Ô/ˆIÝ& x°¸AÐ`ÀaÐQYÐ`Ð`Ð]_Ð`Ð`ˆŒØ!�> )Ð@Ð@°TÐ@¸RÐ@Ð@ˆŒõ &ð 
ØØØ%Ø#ØØ!ØØØ)Ø#9ð
ð 
ð ð
ð 
ˆõ ”m W W¨Y¸
Ñ%CÔ%CÐDˆŒØ#Ô,ˆÔØAÐA¨tÔ/@ÐAÑAÔAˆŒØ#œNˆÔØ ,ˆÔõ 0¸+ÐFÑFÔFˆÔØÔ*¨TÔ-=×-GÒ-GÑ-IÔ-IÑIˆØð 	0å*ØØÔ%Øðð !Øðð ð ðð ˆDŒNð ,˜^¨DÔ,AÐHÐHÀRÐHÐHˆDŒNÝœ™œˆDŒIˆIõ +ØØÔ%Øðð !Øðð ð ðð ˆDŒNõ  œ[™]œ]ˆDŒNØ!˜	¨$Ð/Ñ/Ô/ˆDŒIØ(3ÐF•r”z !‘}”}�}½¼¹¼ˆŒØNYÐ\]ÊoÈo�& Ô!6¸ÐJÐJÀrÐJÐJÐJÕceÔcnÑcpÔcpˆŒå! $Ñ'Ô'Ð'Ð'Ð'rZ   Úreturnc                 óR  — | j         | j        g}|                     | j        ¦  «         |                     | j        | j        | j        | j        g¦  «         |                     t          j	        ¦   «         t          j
        | j        ¦  «        | j        g¦  «         t          j        |Ž S )zqConvert model to sequential form.

        Returns:
            Sequential module containing all layers.
        )re   rf   Úextendrh   rF   ro   rp   rr   ra   rs   ÚDropoutrC   ru   rg   )rv   Úlayerss     rX   Úas_sequentialzMobileNetV3.as_sequential¡   s„   € ð ”. $¤(Ð+ˆØ�Š�d”kÑ"Ô"Ð"Ø�Š�tÔ'¨¬¸¼ÈÌÐSÑTÔTÐTØ�Š•r”z‘|”|¥R¤Z°´Ñ%?Ô%?ÀÄÐQÑRÔRÐRÝŒ}˜fÐ%Ð%rZ   Úcoarsec                 ó,   — t          d|rdnd¬¦  «        S )z"Group parameters for optimization.z^conv_stem|bn1z^blocks\.(\d+)z^blocks\.(\d+)\.(\d+))Ústemrh   )Údict)rv   r‚   s     rX   Úgroup_matcherzMobileNetV3.group_matcher­   s-   € õ Ø"Ø(.ÐLÐ$Ð$Ð4Lð
ñ 
ô 
ð 	
rZ   Úenablec                 ó   — || _         dS ©z)Enable or disable gradient checkpointing.N©rd   ©rv   r‡   s     rX   Úset_grad_checkpointingz"MobileNetV3.set_grad_checkpointingµ   ó   € ð #)ˆÔÐÐrZ   c                 ó   — | j         S )zGet the classifier head.)ru   )rv   s    rX   Úget_classifierzMobileNetV3.get_classifierº   s   € ð ŒÐrZ   c                 óÔ  — t          | ¦  «        }|| _        t          |¬¦  «        | _        | j                             | j        ¦  «         |rt          j        d¦  «        nt          j        ¦   «         | _	        | j	                             | j        ¦  «         |dk    rt          | j        |fi |¤Žnt          j        ¦   «         | _        | j                             | j        ¦  «         dS )zŸReset the classifier head.

        Args:
            num_classes: Number of classes for new classifier.
            global_pool: Global pooling type.
        r[   r   r   N)r   r5   r   rF   ÚtrainÚtrainingra   rs   rq   rt   r   rm   ru   )rv   r5   rF   rw   s       rX   Úreset_classifierzMobileNetV3.reset_classifier¿   sÍ   € õ ˜dÑ#Ô#ˆØ&ˆÔå/¸+ÐFÑFÔFˆÔØÔ×Ò˜tœ}Ñ-Ô-Ð-Ø(3ÐF•r”z !‘}”}�}½¼¹¼ˆŒØŒ×Ò˜4œ=Ñ)Ô)Ð)ØNYÐ\]ÊoÈo�& Ô!6¸ÐJÐJÀrÐJÐJÐJÕceÔcnÑcpÔcpˆŒØŒ×Ò˜dœmÑ,Ô,Ð,Ð,Ð,rZ   ÚNCHWÚxÚindicesÚnormÚ
stop_earlyÚ
output_fmtÚintermediates_onlyÚextra_blocksc                 ó   ‡ — |dv s
J d¦   «         ‚|r|s
J d¦   «         ‚g }|r)t          t          ‰ j        ¦  «        dz   |¦  «        \  }	}
n@t          t          ‰ j        ¦  «        |¦  «        \  }	}
ˆ fd„|	D ¦   «         }	‰ j        |
         }
d}‰                      |¦  «        }‰                      |¦  «        }||	v r|                     |¦  «         t          j         	                    ¦   «         s|s‰ j        }n‰ j        d|
…         }t          |d¬¦  «        D ]_\  }}‰ j        r/t          j         	                    ¦   «         st          ||¦  «        }n ||¦  «        }||	v r|                     |¦  «         Œ`|r|S ||fS )	aa   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
            extra_blocks: Include outputs of all blocks and head conv in output, does not align with feature_info
        Returns:

        )r”   zOutput shape must be NCHW.z/Must use intermediates_only for early stopping.r   c                 ó*   •— g | ]}‰j         |         ‘ŒS rT   )rk   )rV   Úirv   s     €rX   rY   z5MobileNetV3.forward_intermediates.<locals>.<listcomp>ï   s    ø€ ÐEÐEÐE°1˜DœO¨AÔ.ÐEÐEÐErZ   r   N)Ústart)r$   Úlenrh   rk   re   rf   ÚappendÚtorchÚjitÚis_scriptingÚ	enumeraterd   r%   )rv   r•   r–   r—   r˜   r™   rš   r›   ÚintermediatesÚtake_indicesÚ	max_indexÚfeat_idxrh   Úblks   `             rX   Úforward_intermediatesz!MobileNetV3.forward_intermediatesÐ   sÇ  ø€ ð. ˜YÐ&Ð&Ð&Ð(DÑ&Ô&Ð&Øð 	YØ%ÐXÐXÐ'XÑXÔXÐ%ØˆØð 	3Ý&:½3¸t¼{Ñ;KÔ;KÈaÑ;OÐQXÑ&YÔ&YÑ#ˆL˜)˜)å&:½3¸t¼Ñ;OÔ;OÐQXÑ&YÔ&YÑ#ˆL˜)ØEÐEÐEÐE¸ÐEÑEÔEˆLØœ¨	Ô2ˆIð ˆØ�NŠN˜1ÑÔˆØ�HŠH�Q‰KŒKˆØ�|Ð#Ð#Ø× Ò  Ñ#Ô#Ð#åŒ9×!Ò!Ñ#Ô#ð 	-¨:ð 	-Ø”[ˆFˆFà”[  ) Ô,ˆFÝ& v°QÐ7Ñ7Ô7ð 	(ð 	(‰MˆH�cØÔ&ð ­u¬y×/EÒ/EÑ/GÔ/Gð Ý" 3¨Ñ*Ô*��à�C˜‘F”F�Ø˜<Ð'Ð'Ø×$Ò$ QÑ'Ô'Ð'øàð 	!Ø Ð à�-ÐÐrZ   r   Ú
prune_normÚ
prune_headc                 ó  — |r)t          t          | j        ¦  «        dz   |¦  «        \  }}n2t          t          | j        ¦  «        |¦  «        \  }}| j        |         }| j        d|…         | _        |t          | j        ¦  «        k     r0t	          j        ¦   «         | _        t	          j        ¦   «         | _        |rFt	          j        ¦   «         | _        t	          j        ¦   «         | _        |                      dd¦  «         |S )a~  Prune layers not required for specified intermediates.

        Args:
            indices: Indices of intermediate layers to keep.
            prune_norm: Whether to prune normalization layer.
            prune_head: Whether to prune the classifier head.
            extra_blocks: Include outputs of all blocks.

        Returns:
            List of indices that were kept.
        r   Nr   r1   )	r$   r    rh   rk   ra   rq   ro   rp   r“   )rv   r–   r¬   r­   r›   r§   r¨   s          rX   Úprune_intermediate_layersz%MobileNetV3.prune_intermediate_layers
  sÜ   € ð$ ð 	3Ý&:½3¸t¼{Ñ;KÔ;KÈaÑ;OÐQXÑ&YÔ&YÑ#ˆL˜)˜)å&:½3¸t¼Ñ;OÔ;OÐQXÑ&YÔ&YÑ#ˆL˜)Øœ¨	Ô2ˆIØ”k * 9 *Ô-ˆŒØ•s˜4œ;Ñ'Ô'Ò'Ð'Ýœ[™]œ]ˆDŒNÝœ[™]œ]ˆDŒNØð 	)Ýœ[™]œ]ˆDŒNÝœ[™]œ]ˆDŒNØ×!Ò! ! RÑ(Ô(Ð(ØÐrZ   c                 óþ   — |                       |¦  «        }|                      |¦  «        }| j        r6t          j                             ¦   «         st          | j        |d¬¦  «        }n|                      |¦  «        }|S )z’Forward pass through feature extraction layers.

        Args:
            x: Input tensor.

        Returns:
            Feature tensor.
        T)rt   )re   rf   rd   r¢   r£   r¤   r%   rh   ©rv   r•   s     rX   Úforward_featureszMobileNetV3.forward_features+  sp   € ð �NŠN˜1ÑÔˆØ�HŠH�Q‰KŒKˆØÔ"ð 	­5¬9×+AÒ+AÑ+CÔ+Cð 	Ý˜tœ{¨A°tÐ<Ñ<Ô<ˆAˆAà—’˜A‘”ˆAØˆrZ   Ú
pre_logitsc                 ó^  — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }| j        dk    r!t          j        || j        | j        ¬¦  «        }|r|S |  	                    |¦  «        S )zàForward pass through classifier head.

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

        Returns:
            Classification logits or features.
        r2   )Úpr’   )
rF   ro   rp   rr   rt   rC   ÚFÚdropoutr’   ru   )rv   r•   r³   s      rX   Úforward_headzMobileNetV3.forward_head<  s�   € ð ×Ò˜QÑÔˆØ�NŠN˜1ÑÔˆØ�NŠN˜1ÑÔˆØ�IŠI�a‰LŒLˆØ�LŠL˜‰OŒOˆØŒ>˜BÒÐÝ”	˜!˜tœ~¸¼ÐFÑFÔFˆAØð 	ØˆHØ�Š˜qÑ!Ô!Ð!rZ   c                 óZ   — |                       |¦  «        }|                      |¦  «        }|S )zoForward pass.

        Args:
            x: Input tensor.

        Returns:
            Output logits.
        )r²   r¸   r±   s     rX   ÚforwardzMobileNetV3.forwardQ  s/   € ð ×!Ò! !Ñ$Ô$ˆØ×Ò˜aÑ Ô ˆØˆrZ   ©F©T)r3   )NFFr”   FF)r   FTF)%Ú__name__Ú
__module__Ú__qualname__Ú__doc__r   r   ÚintÚboolÚstrr   r   r   Úfloatr`   ra   rg   r�   r¢   r£   Úignorer   r   r†   rŒ   ÚModuler�   r“   ÚTensorr
   r   r	   r«   r¯   r²   r¸   rº   Ú__classcell__©r{   s   @rX   r*   r*      s  ø€ € € € € ðð ð$  $ØØØ"Ø $Ø"Ø#ØØ-1Ø.2Ø,0Ø,0Ø $Ø%3Ø!Ø$&Ø6:Ø$ØØð-r(ð r(à!ðr(ð ðr(ð ð	r(ð
 ðr(ð ðr(ð ðr(ð ðr(ð ðr(ð ðr(ð   	Ô*ðr(ð ! Ô+ðr(ð ˜yÔ)ðr(ð ˜yÔ)ðr(ð ðr(ð  #ð!r(ð" ð#r(ð$ "ð%r(ð& %-¨U¤Oð'r(ð( ð)r(ð r(ð r(ð r(ð r(ð r(ðh
&˜rœ}ð 
&ð 
&ð 
&ð 
&ð „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-ð -¨Cð -¸cð -Èdð -ð -ð -ð -ð( 8<ØØ$Ø$Ø',Ø!&ð8 ð 8 àŒ|ð8 ð ˜e C¨¨c¬ NÔ3Ô4ð8 ð ð	8 ð
 ð8 ð ð8 ð !%ð8 ð ð8 ð 
ˆt�E”LÔ! 5¨¬°t¸E¼LÔ7IÐ)IÔ#JÐJÔ	Kð8 ð 8 ð 8 ð 8 ðx ./Ø$Ø#Ø!&ðð à˜3  S¤	˜>Ô*ðð ðð ð	ð
 ðð 
ˆcŒðð ð ð ðB %¤,ð °5´<ð ð ð ð ð""ð "˜eœlð "¸ð "ÈÌð "ð "ð "ð "ð*˜œð ¨%¬,ð ð ð ð ð ð ð ð rZ   c            $       ó@  ‡ — e Zd ZdZddddddded	d
d
d
d
ddd
d
d
fdedeedf         dededede	dede
dede	dee         dee         dee         dee         dededee         f"ˆ fd„Zej        j        d$de	d d
fd!„¦   «         Zd"ej        d eej                 fd#„Zˆ xZS )%r+   z­MobileNetV3 Feature Extractor.

    A work-in-progress feature extraction module for MobileNet-V3 to use as a backbone for segmentation
    and object detection models.
    )r   r   rK   r.   é   Ú
bottleneckr.   r/   FrP   r1   TNr2   r4   Úout_indices.Úfeature_locationr6   r7   r8   rQ   r<   rB   rA   r=   r>   r?   r@   rC   rD   rE   c                 óú  •— t          ¦   «                              ¦   «          ||dœ}|pt          j        }|pt          j        }|pt
          }|| _        || _        d| _        |s |	|¦  «        }t          ||dfd|dœ|¤Ž| _
         ||fi |¤Ž| _         |d¬¦  «        | _        t          d|||	|
|||||||dœ|¤Ž}t          j         |||¦  «        Ž | _        t!          |j        |¦  «        | _        d	„ | j                             ¦   «         D ¦   «         | _        t+          | ¦  «         d
| _        |dk    rD| j                             d¬¦  «        }t/          ||                      ¦   «         ¦  «        | _        d
S d
S )až  Initialize MobileNetV3Features.

        Args:
            block_args: Arguments for blocks of the network.
            out_indices: Output from stages at indices.
            feature_location: Location of feature before/after each block, must be in ['bottleneck', 'expansion'].
            in_chans: Number of input image channels.
            stem_size: Number of output channels of the initial stem convolution.
            fix_stem: If True, don't scale stem by round_chs_fn.
            output_stride: Output stride of the network.
            pad_type: Type of padding to use for convolution layers.
            round_chs_fn: Callable to round number of filters based on depth multiplier.
            se_from_exp: If True, calculate SE channel reduction from expanded mid channels.
            act_layer: Type of activation layer.
            norm_layer: Type of normalization layer.
            aa_layer: Type of anti-aliasing layer.
            se_layer: Type of Squeeze-and-Excite layer.
            drop_rate: Dropout rate.
            drop_path_rate: Stochastic depth rate.
            layer_scale_init_value: Enable layer scale on compatible blocks if not None.
        rH   Fr.   rK   rL   Tr^   )rQ   r<   rB   rA   r=   r>   r?   r@   rD   rE   rÎ   c                 ó,   — i | ]}|d          |d         “ŒS )rS   ÚindexrT   rU   s     rX   ú
<dictcomp>z0MobileNetV3Features.__init__.<locals>.<dictcomp>²  s"   € Ð]Ð]Ð]¸!˜q œz¨1¨W¬:Ð]Ð]Ð]rZ   NrÌ   )ÚmoduleÚ	hook_type)ÚkeysrT   )r_   r`   ra   rb   rc   r   r6   rC   rd   r   re   rf   Úact1r   rg   rh   r"   ri   rj   Ú	get_dictsÚ_stage_out_idxr   Úfeature_hooksr#   Únamed_modules)rv   r4   rÍ   rÎ   r6   r7   r8   rQ   r<   rB   rA   r=   r>   r?   r@   rC   rD   rE   rI   rJ   rw   ry   Úhooksr{   s                          €rX   r`   zMobileNetV3Features.__init__f  sÐ  ø€ õV 	‰Œ×ÒÑÔÐØ¨Ð/Ð/ˆØÐ(¥¤ˆ	ØÐ1¥2¤>ˆ
ØÐ,�}ˆØ ˆŒØ"ˆŒØ"'ˆÔð ð 	0Ø$˜ YÑ/Ô/ˆIÝ& x°¸AÐ`ÀaÐQYÐ`Ð`Ð]_Ð`Ð`ˆŒØ�:˜iÐ.Ð.¨2Ð.Ð.ˆŒØ�I dÐ+Ñ+Ô+ˆŒ	õ &ð 
Ø'ØØ%Ø#ØØ!ØØØ)Ø#9Ø-ð
ð 
ð ð
ð 
ˆõ ”m W W¨Y¸
Ñ%CÔ%CÐDˆŒÝ'¨Ô(8¸+ÑFÔFˆÔØ]Ð]¸tÔ?P×?ZÒ?ZÑ?\Ô?\Ð]Ñ]Ô]ˆÔå! $Ñ'Ô'Ð'ð "ˆÔØ˜|Ò+Ð+ØÔ%×/Ò/Ð5LÐ/ÑMÔMˆEÝ!-¨e°T×5GÒ5GÑ5IÔ5IÑ!JÔ!JˆDÔÐÐð ,Ð+rZ   r‡   r|   c                 ó   — || _         dS r‰   rŠ   r‹   s     rX   rŒ   z*MobileNetV3Features.set_grad_checkpointing¼  r�   rZ   r•   c                 ót  — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }| j        €žg }d| j        v r|                     |¦  «         t          | j        ¦  «        D ]g\  }}| j        r/t          j
                             ¦   «         st          ||¦  «        }n ||¦  «        }|dz   | j        v r|                     |¦  «         Œh|S |                      |¦  «         | j                             |j        ¦  «        }t          |                     ¦   «         ¦  «        S )z”Forward pass through feature extraction.

        Args:
            x: Input tensor.

        Returns:
            List of feature tensors.
        Nr   r   )re   rf   rÖ   rÙ   rØ   r¡   r¥   rh   rd   r¢   r£   r¤   r&   Ú
get_outputrI   ÚlistÚvalues)rv   r•   ri   rž   ÚbÚouts         rX   rº   zMobileNetV3Features.forwardÁ  s#  € ð �NŠN˜1ÑÔˆØ�HŠH�Q‰KŒKˆØ�IŠI�a‰LŒLˆØÔÐ%ØˆHØ�DÔ'Ð'Ð'Ø—’ Ñ"Ô"Ð"Ý! $¤+Ñ.Ô.ð 'ð '‘��1ØÔ*ð µ5´9×3IÒ3IÑ3KÔ3Kð Ý" 1 aÑ(Ô(�A�Aà˜˜!™œ�AØ�q‘5˜DÔ/Ð/Ð/Ø—O’O AÑ&Ô&Ð&øØˆOà�KŠK˜‰NŒNˆNØÔ$×/Ò/°´Ñ9Ô9ˆCÝ˜Ÿ
š
™œÑ%Ô%Ð%rZ   r¼   )r½   r¾   r¿   rÀ   r   r   r	   rÁ   rÃ   rÂ   r   r   r   r   rÄ   r`   r¢   r£   rÅ   rŒ   rÇ   r   rº   rÈ   rÉ   s   @rX   r+   r+   _  sá  ø€ € € € € ðð ð ,;Ø$0ØØØ"Ø!#Ø "Ø%3Ø $Ø-1Ø.2Ø,0Ø,0Ø!Ø$&Ø6:ØØð)TKð TKà!ðTKð ˜s C˜xœðTKð "ð	TKð
 ðTKð ðTKð ðTKð ðTKð ðTKð #ðTKð ðTKð   	Ô*ðTKð ! Ô+ðTKð ˜yÔ)ðTKð ˜yÔ)ðTKð  ð!TKð" "ð#TKð$ %-¨U¤Oð%TKð TKð TKð TKð TKð TKðl „YÔð)ð )¨Tð )¸Tð )ð )ð )ñ Ôð)ð&˜œð &¨$¨u¬|Ô*<ð &ð &ð &ð &ð &ð &ð &ð &rZ   FÚvariantÚ
pretrainedr|   c                 óè   — d}t           }d}|                     dd¦  «        rd|v sd|v rd}nd}t          }d	}t          || |f|dk    |d	k    |d
œ|¤Ž}|d	k    rt	          |j        ¦  «        |_        |S )zÝCreate a MobileNetV3 model.

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

    Returns:
        MobileNetV3 model instance.
    r1   NÚfeatures_onlyFÚfeature_cfgÚfeature_clsÚcfg)r5   r9   Ú	head_convr:   r;   rF   Úcls)ræ   Úpretrained_strictÚkwargs_filter)r*   Úpopr+   r   r   Údefault_cfg)rã   rä   ÚkwargsÚfeatures_modeÚ	model_clsrí   Úmodels          rX   Ú_create_mnv3rô   ß  sº   € ð €MÝ€IØ€MØ‡z‚z�/ 5Ñ)Ô)ð "Ø˜FÐ"Ð" m°vÐ&=Ð&=Ø!ˆMˆMàqˆMÝ+ˆIØ!ˆMå ØØØðð $ uÒ,Ø'¨5Ò0Ø#ðð ð ðð €Eð ˜ÒÐÝ7¸Ô8IÑJÔJˆÔØ€LrZ   ç      ð?Úchannel_multiplierc                 ó6  — dgddgdgg d¢dgdgdgg}t          dt          |¦  «        d	t          t          |¬
¦  «        t          t          j        fi t          |¦  «        ¤Žt          |d¦  «        t          t          d¬¦  «        dœ|¤Ž}t          | |fi |¤Ž}|S )ad  Creates a MobileNet-V3 model.

    Ref impl: ?
    Paper: https://arxiv.org/abs/1905.02244

    Args:
        variant: Model variant name.
        channel_multiplier: Multiplier to number of channels per layer.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        MobileNetV3 model instance.
    Úds_r1_k3_s1_e1_c16_nre_noskipÚir_r1_k3_s2_e4_c24_nreÚir_r1_k3_s1_e3_c24_nreúir_r3_k5_s2_e3_c40_se0.25_nre©Úir_r1_k3_s2_e6_c80zir_r1_k3_s1_e2.5_c80zir_r2_k3_s1_e2.3_c80úir_r2_k3_s1_e6_c112_se0.25úir_r3_k5_s2_e6_c160_se0.25Úcn_r1_k1_s1_c960F©Ú
multiplierÚ
hard_swishÚhard_sigmoid)Ú
gate_layer)r4   r:   rB   r>   r=   r@   rT   )
r…   r   r   r   ra   rc   r   r    r   rô   ©rã   rö   rä   rð   Úarch_defÚmodel_kwargsró   s          rX   Ú_gen_mobilenet_v3_rwr	    s×   € ð& 
)Ð)à	!Ð#;Ð<à	(Ð)àNÐNÐNà	%Ð&à	%Ð&à	Ðð€Hõ  ð Ý" 8Ñ,Ô,ØÝ�^Ð8JÐKÑKÔKÝ�2œ>ÐEÐE­_¸VÑ-DÔ-DÐEÐEÝ# F¨LÑ9Ô9Ý�°>ÐBÑBÔBðð ð ðð €Lõ ˜ *Ð=Ð=°Ð=Ð=€EØ€LrZ   Údepth_multiplierÚ
group_sizec                 ó\  — d| v rHd}d| v r!t          |d¦  «        }dgddgdd	gd
gdgdgg}nlt          |d¦  «        }dgddgddgdgdgdgg}nKd}d| v r#t          |d¦  «        }dgddgdgg d¢dgdgdgg}n"t          |d¦  «        }dgdd gd!gg d¢d"gd#gdgg}t          t          d$t          j        t
          ¬%¦  «        }	t          d+t          |||¬&¦  «        |d'|d(k     t          t
          |¬)¦  «        t          t          j        fi t          |¦  «        ¤Ž||	d*œ|¤Ž}
t          | |fi |
¤Ž}|S ),aÛ  Creates a MobileNet-V3 model.

    Ref impl: ?
    Paper: https://arxiv.org/abs/1905.02244

    Args:
        variant: Model variant name.
        channel_multiplier: Multiplier to number of channels per layer.
        depth_multiplier: Depth multiplier for model scaling.
        group_size: Group size for grouped convolutions.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        MobileNetV3 model instance.
    Úsmalli   ÚminimalÚreluÚds_r1_k3_s2_e1_c16zir_r1_k3_s2_e4.5_c24zir_r1_k3_s1_e3.67_c24Úir_r1_k3_s2_e4_c40Úir_r2_k3_s1_e6_c40Úir_r2_k3_s1_e3_c48Úir_r3_k3_s2_e6_c96Úcn_r1_k1_s1_c576r  zds_r1_k3_s2_e1_c16_se0.25_nrezir_r1_k3_s2_e4.5_c24_nrezir_r1_k3_s1_e3.67_c24_nreúir_r1_k5_s2_e4_c40_se0.25zir_r2_k5_s1_e6_c40_se0.25zir_r2_k5_s1_e3_c48_se0.25zir_r3_k5_s2_e6_c96_se0.25r0   Úds_r1_k3_s1_e1_c16Úir_r1_k3_s2_e4_c24Úir_r1_k3_s1_e3_c24Úir_r3_k3_s2_e3_c40rü   Úir_r2_k3_s1_e6_c112Úir_r3_k3_s2_e6_c160r   Úds_r1_k3_s1_e1_c16_nrerù   rú   rû   rþ   rÿ   r  )r  Úforce_act_layerÚrd_round_fn)r
  r  r/   ç      è?r  )r4   r9   r7   r8   rB   r>   r=   r@   rT   )r    r   r   ra   rb   r   r…   r   rc   r   rô   )rã   rö   r
  r  rä   rð   r9   r=   r  r@   r  ró   s               rX   Ú_gen_mobilenet_v3r!  1  s  € ð0 �'ÐÐØˆØ˜ÐÐÝ)¨&°&Ñ9Ô9ˆIð &Ð&à'Ð)@ÐAà%Ð';Ð<à%Ð&à%Ð&à#Ð$ðˆHˆHõ *¨&°,Ñ?Ô?ˆIð 1Ð1à+Ð-HÐIà,Ð.IÐJà,Ð-à,Ð-à#Ð$ðˆHˆHð ˆØ˜ÐÐÝ)¨&°&Ñ9Ô9ˆIð &Ð&à%Ð';Ð<à%Ð&àVÐVÐVà&Ð'à&Ð'à#Ð$ðˆHˆHõ" *¨&°,Ñ?Ô?ˆIð *Ð*à)Ð+CÐDà0Ð1àVÐVÐVà-Ð.à-Ð.à#Ð$ðˆHõ  •}°ÕQSÔQXÕftÐuÑuÔu€HÝð 
Ý" 8Ð>NÐ[eÐfÑfÔfØ!ØØ# dÒ*Ý�^Ð8JÐKÑKÔKÝ�2œ>ÐEÐE­_¸VÑ-DÔ-DÐEÐEØØð
ð 
ð ð
ð 
€Lõ ˜ *Ð=Ð=°Ð=Ð=€EØ€LrZ   c                 ó  — |                       d¦  «        d         }|dv rd}dgddgdd	gd
dgddgg d¢dgg}nC|dk    rd}dgddgddgddgddgg d¢dgg}n%|dk    rd}dgddgd d!gd"d#gd$d%gg d&¢d'gg}nt          ‚t          t          |d(¬)¦  «        }t          t          d*|¬+¦  «        }t          |d,¦  «        }	t          d0t          |¦  «        d-d.||d.t          t          j	        fi t          |¦  «        ¤Ž|	|d/œ	|¤Ž}
t          | |fi |
¤Ž}|S )1aÙ  FBNetV3 model generator.

    Paper: `FBNetV3: Joint Architecture-Recipe Search using Predictor Pretraining`
        - https://arxiv.org/abs/2006.02049
    FIXME untested, this is a preliminary impl of some FBNet-V3 variants.

    Args:
        variant: Model variant name.
        channel_multiplier: Channel width multiplier.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        MobileNetV3 model instance.
    Ú_éÿÿÿÿ)Úará   r/   Úds_r2_k3_s1_e1_c16Úir_r1_k5_s2_e4_c24Úir_r3_k5_s1_e2_c24zir_r1_k5_s2_e5_c40_se0.25zir_r4_k5_s1_e3_c40_se0.25Úir_r1_k5_s2_e5_c72Úir_r4_k3_s1_e3_c72zir_r1_k3_s1_e5_c120_se0.25zir_r5_k5_s1_e3_c120_se0.25)zir_r1_k3_s2_e6_c184_se0.25zir_r5_k5_s1_e4_c184_se0.25zir_r1_k5_s1_e6_c224_se0.25Úcn_r1_k1_s1_c1344Údé   Úir_r1_k3_s2_e5_c24Úir_r5_k3_s1_e2_c24r  zir_r4_k3_s1_e3_c40_se0.25Úir_r1_k3_s2_e5_c72zir_r1_k3_s1_e5_c128_se0.25zir_r6_k5_s1_e3_c128_se0.25)zir_r1_k3_s2_e6_c208_se0.25zir_r5_k5_s1_e5_c208_se0.25zir_r1_k5_s1_e6_c240_se0.25Úcn_r1_k1_s1_c1440ÚgrP   Úds_r3_k3_s1_e1_c24Úir_r1_k5_s2_e4_c40Úir_r4_k5_s1_e2_c40zir_r1_k5_s2_e4_c56_se0.25zir_r4_k5_s1_e3_c56_se0.25Úir_r1_k5_s2_e5_c104Úir_r4_k3_s1_e3_c104zir_r1_k3_s1_e5_c160_se0.25zir_r8_k5_s1_e3_c160_se0.25)zir_r1_k3_s2_e6_c264_se0.25zir_r6_k5_s1_e5_c264_se0.25zir_r2_k5_s1_e6_c288_se0.25Úcn_r1_k1_s1_c1728çffffffî?)r  Úround_limitr  )r  r  r  iÀ  F)	r4   r9   r:   r7   rB   rA   r>   r=   r@   rT   )ÚsplitÚNotImplementedr   r   r   r    r…   r   ra   rc   r   rô   )rã   rö   rä   rð   Úvlr7   r  rB   r@   r=   r  ró   s               rX   Ú_gen_fbnetv3r>  ¡  sÏ  € ð  
�Š�sÑ	Ô	˜BÔ	€BØ	ˆZÐÐØˆ	à!Ð"Ø!Ð#7Ð8Ø(Ð*EÐFØ!Ð#7Ð8Ø)Ð+GÐHØfÐfÐfØ Ð!ð
ˆˆð 
ˆsŠˆØˆ	à!Ð"Ø!Ð#7Ð8Ø(Ð*EÐFØ!Ð#7Ð8Ø)Ð+GÐHØfÐfÐfØ Ð!ð
ˆˆð 
ˆsŠˆØˆ	à!Ð"Ø!Ð#7Ð8Ø(Ð*EÐFØ"Ð$9Ð:Ø)Ð+GÐHØfÐfÐfØ Ð!ð
ˆˆõ ÐÝ�>Ð6HÐVZÐ[Ñ[Ô[€LÝ•}°È\ÐZÑZÔZ€HÝ! &¨,Ñ7Ô7€IÝð Ý" 8Ñ,Ô,ØØØØ!ØÝ�2œ>ÐEÐE­_¸VÑ-DÔ-DÐEÐEØØðð ð ðð €Lõ ˜ *Ð=Ð=°Ð=Ð=€EØ€LrZ   c                 óH  — dgdgdgddgdgdgg}t          dt          |¦  «        dt          t          |¬	¦  «        t          t          j        fi t          |¦  «        ¤Žt          |d
¦  «        t          t          dt          j	        ¬¦  «        ddœ|¤Ž}t          | |fi |¤Ž}|S )aÆ  LCNet model generator.

    Essentially a MobileNet-V3 crossed with a MobileNet-V1

    Paper: `PP-LCNet: A Lightweight CPU Convolutional Neural Network` - https://arxiv.org/abs/2109.15099

    Args:
        variant: Model variant name.
        channel_multiplier: Multiplier to number of channels per layer.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        MobileNetV3 model instance.
    Údsa_r1_k3_s1_c32Údsa_r2_k3_s2_c64Údsa_r2_k3_s2_c128Údsa_r1_k3_s2_c256Údsa_r1_k5_s1_c256Údsa_r4_k5_s1_c256zdsa_r2_k5_s2_c512_se0.25r/   r  r  r  )r  r  r0   )r4   r7   rB   r>   r=   r@   r9   rT   )r…   r   r   r   ra   rc   r   r    r   rb   rô   r  s          rX   Ú
_gen_lcnetrF  è  s×   € ð$ 
Ðà	Ðà	Ðà	Ð1Ð2à	Ðà	#Ð$ð€Hõ ð 	Ý" 8Ñ,Ô,ØÝ�^Ð8JÐKÑKÔKÝ�2œ>ÐEÐE­_¸VÑ-DÔ-DÐEÐEÝ# F¨LÑ9Ô9Ý�°>ÕSUÔSZÐ[Ñ[Ô[Øð	ð 	ð ð	ð 	€Lõ ˜ *Ð=Ð=°Ð=Ð=€EØ€LrZ   c                 ó¬  — d}d| v r\d}d| v r"d}t          |d¦  «        }dgdd	gg d
¢g d¢dgg}	n·d| v r"d}t          |d¦  «        }dgddgg d¢g d¢dgg}	n‘J d| › d�¦   «         ‚d}d| v r#d}t          |d¦  «        }ddgddgg d¢g d¢dgg}	nZd| v r"d}t          |d¦  «        }dgdd	gg d¢g d ¢dgg}	n4d| v r"d}t          |d¦  «        }dgddgg d!¢g d"¢dgg}	nJ d| › d�¦   «         ‚t          d(t          |	|¬#¦  «        dd$|||d%k     t          t          |¬&¦  «        t          t
          j        fi t          |¦  «        ¤Ž||d'œ
|¤Ž}
t          | |fi |
¤Ž}|S ))a�  Creates a MobileNet-V4 model.

    Paper: https://arxiv.org/abs/2404.10518

    Args:
        variant: Model variant name.
        channel_multiplier: Multiplier to number of channels per layer.
        group_size: Group size for grouped convolutions.
        pretrained: Load pretrained weights.
        **kwargs: Additional model arguments.

    Returns:
        MobileNetV3 model instance.
    r0   Úhybridgñhãˆµøä>ÚmediumrP   r  Úer_r1_k3_s2_e4_c48Úuir_r1_a3_k5_s2_e4_c80Úuir_r1_a3_k3_s1_e2_c80)Úuir_r1_a3_k5_s2_e6_c160Úuir_r1_a0_k0_s1_e2_c160Úuir_r1_a3_k3_s1_e4_c160Úuir_r1_a3_k5_s1_e4_c160Úmqa_r1_k3_h4_s1_v2_d64_c160rO  rQ  Úuir_r1_a3_k0_s1_e4_c160rQ  rO  rQ  rR  )Úuir_r1_a5_k5_s2_e6_c256Úuir_r1_a5_k5_s1_e4_c256Úuir_r2_a3_k5_s1_e4_c256Úuir_r1_a0_k0_s1_e2_c256Úuir_r1_a3_k5_s1_e2_c256rV  Úuir_r1_a0_k0_s1_e4_c256Úmqa_r1_k3_h4_s1_d64_c256Úuir_r1_a3_k0_s1_e4_c256rY  rT  rY  Úuir_r1_a5_k0_s1_e4_c256rY  r[  r   Úlarger-  ÚgeluÚuir_r1_a3_k5_s2_e4_c96Úuir_r1_a3_k3_s1_e4_c96)Úuir_r1_a3_k5_s2_e4_c192Úuir_r3_a3_k3_s1_e4_c192Úuir_r1_a3_k5_s1_e4_c192Úuir_r2_a5_k3_s1_e4_c192Úmqa_r1_k3_h8_s1_v2_d48_c192Úuir_r1_a5_k3_s1_e4_c192rd  re  rd  re  rd  Úuir_r1_a3_k0_s1_e4_c192)Úuir_r4_a5_k5_s2_e4_c512Úuir_r1_a5_k0_s1_e4_c512Úuir_r1_a5_k3_s1_e4_c512Úuir_r2_a5_k0_s1_e4_c512ri  Úuir_r1_a5_k5_s1_e4_c512Úmqa_r1_k3_h8_s1_d64_c512rh  rl  rh  rl  rh  rl  rh  FzUnknown variant ú.Nr  Úcn_r1_k3_s2_e1_c32Úcn_r1_k1_s1_e1_c32Úcn_r1_k3_s2_e1_c96Úcn_r1_k1_s1_e1_c64)Úuir_r1_a5_k5_s2_e3_c96Úuir_r4_a0_k3_s1_e2_c96Úuir_r1_a3_k0_s1_e4_c96)Úuir_r1_a3_k3_s2_e6_c128Úuir_r1_a5_k5_s1_e4_c128Úuir_r1_a0_k5_s1_e4_c128Úuir_r1_a0_k5_s1_e3_c128Úuir_r2_a0_k3_s1_e4_c128)rM  Úuir_r2_a3_k3_s1_e4_c160rP  rO  rR  rN  rR  )	rS  rT  rU  rX  rZ  rW  rT  Úuir_r2_a0_k0_s1_e4_c256Úuir_r1_a5_k0_s1_e2_c256)r`  ra  rb  Úuir_r5_a5_k3_s1_e4_c192rf  )rg  rh  ri  rj  ri  rk  Úuir_r3_a5_k0_s1_e4_c512)r  Trõ   r  )
r4   r:   r;   r9   r7   r8   rB   r>   r=   rE   rT   )	r    r…   r   r   r   ra   rc   r   rô   )rã   rö   r  rä   rð   r9   rE   r7   r=   r  r  ró   s               rX   Ú_gen_mobilenet_v4r    s  € ð* €LØ�7ÐÐØ!%ÐØ�wÐÐØˆIÝ)¨&°&Ñ9Ô9ˆIð )ðð
 -Ø,ðð
ð ð ðð ð ð& 'ððY/ˆHˆHð` ˜ÐÐØˆIÝ)¨&°&Ñ9Ô9ˆIð )ðð
 -Ø,ðð
ð ð ðð ð ð$ 'ððW.ˆHˆHð` 8Ð7¨WÐ7Ð7Ð7Ñ7Ô7�5à!%ÐØ�gÐÐØˆIÝ)¨&°&Ñ9Ô9ˆIð )Ø(ðð )Ø(ðð
ð ð ðð ð ð 'ðð5ˆHˆHð< ˜Ð Ð ØˆIÝ)¨&°&Ñ9Ô9ˆIð )ðð
 -Ø,ðð
ð ð ð
ð 
ð 
ð 'ððC$ˆHˆHðJ ˜ÐÐØˆIÝ)¨&°&Ñ9Ô9ˆIð )ðð
 -Ø,ðð
ð ð ð	ð 	ð 	ð 'ðð=!ˆHˆHðF 8Ð7¨WÐ7Ð7Ð7Ñ7Ô7�5åð Ý" 8¸
ÐCÑCÔCØØØ!ØØ# cÒ)Ý�^Ð8JÐKÑKÔKÝ�2œ>ÐEÐE­_¸VÑ-DÔ-DÐEÐEØØ5ðð ð ðð €Lõ ˜ *Ð=Ð=°Ð=Ð=€EØ€LrZ   r1   Úurlc                 ó6   — | dddddt           t          dddd	œ|¥S )
z»Create default configuration dictionary.

    Args:
        url: Model weight URL.
        **kwargs: Additional configuration options.

    Returns:
        Configuration dictionary.
    r-   )r.   éà   r‚  )é   rƒ  g      ì?Úbilinearre   ru   z
apache-2.0)r€  r5   Ú
input_sizeÚ	pool_sizeÚcrop_pctÚinterpolationÚmeanÚstdÚ
first_convru   Úlicense)r   r   )r€  rð   s     rX   Ú_cfgr�    s:   € ð  4°}ÐSYØ¨JÝ%Õ.BØ!°Øðð ð
 $*ðð rZ   zmobilenetv3_large_075.untrained)r€  zmobilenetv3_large_100.ra_in1kÚbicubiczvhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_large_100_ra-f55367f5.pthztimm/)rˆ  r€  Ú	hf_hub_idz)mobilenetv3_large_100.ra4_e3600_r224_in1kr9  )r.   é   r�  )r�  rˆ  r‰  rŠ  r‡  Útest_input_sizeÚtest_crop_pctz(mobilenetv3_large_100.miil_in21k_ft_in1kr„  )r2   r2   r2   )rõ   rõ   rõ   z+https://github.com/Alibaba-MIIL/ImageNet21KzarXiv:2104.10972v4z€https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/mobilenetv3_large_100_1k_miil_78_0-66471c13.pth)rˆ  r‰  rŠ  Ú
origin_urlÚ	paper_idsr€  r�  z mobilenetv3_large_100.miil_in21kz~https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/mobilenetv3_large_100_in21k_miil-d71cc17b.pthiÕ+  )r€  r�  r“  r”  rˆ  r‰  rŠ  r5   z*mobilenetv3_large_150d.ra4_e3600_r256_in1k)é   r•  )r.   é@  r–  )r�  r‰  rŠ  r…  r‡  r†  r‘  r’  zmobilenetv3_small_050.lamb_in1kzyhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_small_050_lambc-4b7bbe87.pth)r€  r�  rˆ  zmobilenetv3_small_075.lamb_in1kzyhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_small_075_lambc-384766db.pthzmobilenetv3_small_100.lamb_in1kzxhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_small_100_lamb-266a294c.pthzmobilenetv3_rw.rmsp_in1kzmhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_100-35495452.pthztf_mobilenetv3_large_075.in1kzvhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_large_075-150ee8b0.pth)r€  r�  r‰  rŠ  ztf_mobilenetv3_large_100.in1kzvhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_large_100-427764d5.pthz%tf_mobilenetv3_large_minimal_100.in1kz~https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_large_minimal_100-8596ae28.pthztf_mobilenetv3_small_075.in1kzvhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_small_075-da427f52.pthztf_mobilenetv3_small_100.in1kzvhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_small_100-37f49e2b.pthz%tf_mobilenetv3_small_minimal_100.in1kz~https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_small_minimal_100-922a7843.pthzfbnetv3_b.ra2_in1kzkhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetv3_b_224-ead5d2a1.pth)r€  r�  r‘  r‡  zfbnetv3_d.ra2_in1kzkhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetv3_d_224-c98bce42.pthzfbnetv3_g.ra2_in1kzkhttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetv3_g_240-0b1df83b.pth)r.   éð   r—  )r.   é   r˜  )r€  r�  r…  r‘  r‡  r†  zlcnet_035.untrainedzlcnet_050.ra2_in1kzghttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/lcnet_050-f447553b.pthzlcnet_075.ra2_in1kzghttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/lcnet_075-318cad2c.pthzlcnet_100.ra2_in1kzghttps://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/lcnet_100-a929038c.pthzlcnet_150.untrainedz$mobilenetv4_conv_small_035.untrained)r‰  rŠ  r‘  r’  rˆ  z*mobilenetv4_conv_small_050.e3000_r224_in1k)r�  r‰  rŠ  r‘  r’  rˆ  z&mobilenetv4_conv_small.e2400_r224_in1k)r�  r‘  r’  rˆ  z&mobilenetv4_conv_small.e1200_r224_in1kz&mobilenetv4_conv_small.e3600_r256_in1k)	r�  r‰  rŠ  r…  r†  r‡  r‘  r’  rˆ  z&mobilenetv4_conv_medium.e500_r256_in1k)r�  r…  r†  r‡  r‘  r’  rˆ  z&mobilenetv4_conv_medium.e500_r224_in1k)r�  r‡  r‘  r’  rˆ  z/mobilenetv4_conv_medium.e250_r384_in12k_ft_in1k)r.   é€  r™  )é   rš  )r�  r…  r†  r‡  rˆ  z'mobilenetv4_conv_medium.e180_r384_in12ki-.  )r�  r5   r…  r†  r‡  rˆ  z*mobilenetv4_conv_medium.e180_ad_r384_in12kz'mobilenetv4_conv_medium.e250_r384_in12kz%mobilenetv4_conv_large.e600_r384_in1k)r.   éÀ  r›  z%mobilenetv4_conv_large.e500_r256_in1kz1mobilenetv4_hybrid_medium.e200_r256_in12k_ft_in1kz+mobilenetv4_hybrid_medium.ix_e550_r256_in1kz+mobilenetv4_hybrid_medium.ix_e550_r384_in1kz(mobilenetv4_hybrid_medium.e500_r224_in1kz)mobilenetv4_hybrid_medium.e200_r256_in12k)r�  r5   r…  r†  r‡  r‘  r’  rˆ  z*mobilenetv4_hybrid_large.ix_e600_r384_in1kz'mobilenetv4_hybrid_large.e600_r384_in1kz$mobilenetv4_conv_aa_medium.untrained)r…  r†  r‡  rˆ  z+mobilenetv4_conv_blur_medium.e500_r224_in1kz1mobilenetv4_conv_aa_large.e230_r448_in12k_ft_in1k)é   rœ  )r.   é   r�  z1mobilenetv4_conv_aa_large.e230_r384_in12k_ft_in1k)r.   éà  rž  z(mobilenetv4_conv_aa_large.e600_r384_in1kz)mobilenetv4_conv_aa_large.e230_r384_in12kz'mobilenetv4_hybrid_medium_075.untrained)r‡  rˆ  z&mobilenetv4_hybrid_large_075.untrainedc                 ó"   — t          dd| i|¤Ž}|S )ú MobileNet V3 Úmobilenetv3_large_075r   rä   )r¡  r   ©r!  ©rä   rð   ró   s      rX   r¡  r¡    ó"   € õ Ð]Ð]È
Ð]ÐV\Ð]Ð]€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )r   Úmobilenetv3_large_100rõ   rä   )r¦  rõ   r¢  r£  s      rX   r¦  r¦    ó"   € õ Ð\Ð\ÀzÐ\ÐU[Ð\Ð\€EØ€LrZ   c                 ó$   — t          dd| dœ|¤Ž}|S )r   Úmobilenetv3_large_150dç      ø?g333333ó?)r
  rä   )r©  rª  r¢  r£  s      rX   r©  r©    s%   € õ ÐsÈcÐ^hÐsÐsÐlrÐsÐs€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )r   Úmobilenetv3_small_050ç      à?rä   )r¬  r­  r¢  r£  s      rX   r¬  r¬    r¤  rZ   c                 ó"   — t          dd| i|¤Ž}|S )r   Úmobilenetv3_small_075r   rä   )r¯  r   r¢  r£  s      rX   r¯  r¯    r¤  rZ   c                 ó"   — t          dd| i|¤Ž}|S )r   Úmobilenetv3_small_100rõ   rä   )r±  rõ   r¢  r£  s      rX   r±  r±  &  r§  rZ   c                 óX   — |                      dt          ¦  «         t          dd| i|¤Ž}|S )r   Úbn_epsÚmobilenetv3_rwrõ   rä   )r´  rõ   )Ú
setdefaultr!   r	  r£  s      rX   r´  r´  -  s9   € ð ×Ò�hÕ 1Ñ2Ô2Ð2Ý ÐXÐXÀ:ÐXÐQWÐXÐX€EØ€LrZ   c                 ó„   — |                      dt          ¦  «         |                      dd¦  «         t          dd| i|¤Ž}|S )r   r³  r<   ÚsameÚtf_mobilenetv3_large_075r   rä   )r¸  r   ©rµ  r!   r!  r£  s      rX   r¸  r¸  5  óO   € ð ×Ò�hÕ 1Ñ2Ô2Ð2Ø
×Ò�j &Ñ)Ô)Ð)ÝÐ`Ð`È:Ð`ÐY_Ð`Ð`€EØ€LrZ   c                 ó„   — |                      dt          ¦  «         |                      dd¦  «         t          dd| i|¤Ž}|S )r   r³  r<   r·  Útf_mobilenetv3_large_100rõ   rä   )r¼  rõ   r¹  r£  s      rX   r¼  r¼  >  óO   € ð ×Ò�hÕ 1Ñ2Ô2Ð2Ø
×Ò�j &Ñ)Ô)Ð)ÝÐ_Ð_È*Ð_ÐX^Ð_Ð_€EØ€LrZ   c                 ó„   — |                      dt          ¦  «         |                      dd¦  «         t          dd| i|¤Ž}|S )r   r³  r<   r·  Ú tf_mobilenetv3_large_minimal_100rõ   rä   )r¿  rõ   r¹  r£  s      rX   r¿  r¿  G  óP   € ð ×Ò�hÕ 1Ñ2Ô2Ð2Ø
×Ò�j &Ñ)Ô)Ð)ÝÐgÐgÐR\ÐgÐ`fÐgÐg€EØ€LrZ   c                 ó„   — |                      dt          ¦  «         |                      dd¦  «         t          dd| i|¤Ž}|S )r   r³  r<   r·  Útf_mobilenetv3_small_075r   rä   )rÂ  r   r¹  r£  s      rX   rÂ  rÂ  P  rº  rZ   c                 ó„   — |                      dt          ¦  «         |                      dd¦  «         t          dd| i|¤Ž}|S )r   r³  r<   r·  Útf_mobilenetv3_small_100rõ   rä   )rÄ  rõ   r¹  r£  s      rX   rÄ  rÄ  Y  r½  rZ   c                 ó„   — |                      dt          ¦  «         |                      dd¦  «         t          dd| i|¤Ž}|S )r   r³  r<   r·  Ú tf_mobilenetv3_small_minimal_100rõ   rä   )rÆ  rõ   r¹  r£  s      rX   rÆ  rÆ  b  rÀ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )z FBNetV3-B Ú	fbnetv3_brä   )rÈ  ©r>  r£  s      rX   rÈ  rÈ  k  ó!   € õ ÐFÐF°ÐF¸vÐFÐF€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )z FBNetV3-D Ú	fbnetv3_drä   )rÌ  rÉ  r£  s      rX   rÌ  rÌ  r  rÊ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )z FBNetV3-G Ú	fbnetv3_grä   )rÎ  rÉ  r£  s      rX   rÎ  rÎ  y  rÊ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )z PP-LCNet 0.35Ú	lcnet_035çffffffÖ?rä   )rÐ  rÑ  ©rF  r£  s      rX   rÐ  rÐ  €  ó!   € õ ÐJÐJ°ZÐJÀ6ÐJÐJ€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )z PP-LCNet 0.5Ú	lcnet_050r­  rä   )rÕ  r­  rÒ  r£  s      rX   rÕ  rÕ  ‡  ó!   € õ ÐIÐI°JÐIÀ&ÐIÐI€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )ú PP-LCNet 1.0Ú	lcnet_075r   rä   )rÙ  r   rÒ  r£  s      rX   rÙ  rÙ  Ž  rÓ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )rØ  Ú	lcnet_100rõ   rä   )rÛ  rõ   rÒ  r£  s      rX   rÛ  rÛ  •  rÖ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )z PP-LCNet 1.5Ú	lcnet_150rª  rä   )rÝ  rª  rÒ  r£  s      rX   rÝ  rÝ  œ  rÖ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )ú MobileNet V4 Úmobilenetv4_conv_small_035rÑ  rä   )rà  rÑ  ©r  r£  s      rX   rà  rà  £  ó"   € õ ÐbÐbÈZÐbÐ[aÐbÐb€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )rß  Úmobilenetv4_conv_small_050r­  rä   )rä  r­  rá  r£  s      rX   rä  rä  ª  râ  rZ   c                 ó"   — t          dd| i|¤Ž}|S )rß  Úmobilenetv4_conv_smallrõ   rä   )ræ  rõ   rá  r£  s      rX   ræ  ræ  ±  r¤  rZ   c                 ó"   — t          dd| i|¤Ž}|S )rß  Úmobilenetv4_conv_mediumrõ   rä   )rè  rõ   rá  r£  s      rX   rè  rè  ¸  s"   € õ Ð^Ð^ÈÐ^ÐW]Ð^Ð^€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )rß  Úmobilenetv4_conv_largerõ   rä   )rê  rõ   rá  r£  s      rX   rê  rê  ¿  r¤  rZ   c                 ó"   — t          dd| i|¤Ž}|S )ú MobileNet V4 Hybrid Úmobilenetv4_hybrid_mediumrõ   rä   )rí  rõ   rá  r£  s      rX   rí  rí  Æ  s"   € õ Ð`Ð`È:Ð`ÐY_Ð`Ð`€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )ú MobileNet V4 HybridÚmobilenetv4_hybrid_largerõ   rä   )rð  rõ   rá  r£  s      rX   rð  rð  Í  s"   € õ Ð_Ð_È*Ð_ÐX^Ð_Ð_€EØ€LrZ   c                 ó$   — t          d| ddœ|¤Ž}|S )ú MobileNet V4 w/ AvgPool AA Úmobilenetv4_conv_aa_mediumrõ   r3   ©rä   r?   )ró  rõ   rá  r£  s      rX   ró  ró  Ô  s%   € õ ÐqÈJÐafÐqÐqÐjpÐqÐq€EØ€LrZ   c                 ó$   — t          d| ddœ|¤Ž}|S )z MobileNet V4 Conv w/ Blur AA Úmobilenetv4_conv_blur_mediumrõ   Úblurpcrô  )rö  rõ   rá  r£  s      rX   rö  rö  Û  s%   € õ ÐvÈjÐckÐvÐvÐouÐvÐv€EØ€LrZ   c                 ó$   — t          d| ddœ|¤Ž}|S )rò  Úmobilenetv4_conv_aa_largerõ   r3   rô  )rù  rõ   rá  r£  s      rX   rù  rù  â  s%   € õ ÐpÈ:Ð`eÐpÐpÐioÐpÐp€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )rì  Úmobilenetv4_hybrid_medium_075r   rä   )rû  r   rá  r£  s      rX   rû  rû  é  s#   € õ ÐeÐeÐPZÐeÐ^dÐeÐe€EØ€LrZ   c                 ó"   — t          dd| i|¤Ž}|S )rï  Úmobilenetv4_hybrid_large_075r   rä   )rý  r   rá  r£  s      rX   rý  rý  ð  s"   € õ ÐdÐdÈzÐdÐ]cÐdÐd€EØ€LrZ   )Úmobilenetv3_large_100_miilÚ mobilenetv3_large_100_miil_in21kr»   )rõ   F)rõ   rõ   NF)rõ   NF)r1   )irÀ   Ú	functoolsr   Útypingr   r   r   r   r   r	   r
   r¢   Útorch.nnra   Útorch.nn.functionalÚ
functionalr¶   Ú	timm.datar   r   r   r   Útimm.layersr   r   r   r   r   r   r   Ú_builderr   r   Ú_efficientnet_blocksr   Ú_efficientnet_builderr   r   r   r   r   r   r    r!   Ú	_featuresr"   r#   r$   Ú_manipulater%   r&   Ú	_registryr'   r(   r)   Ú__all__rÆ   r*   r+   rÃ   rÂ   rô   rÄ   r	  rÁ   r!  r>  rF  r  r�  Údefault_cfgsr¡  r¦  r©  r¬  r¯  r±  r´  r¸  r¼  r¿  rÂ  rÄ  rÆ  rÈ  rÌ  rÎ  rÐ  rÕ  rÙ  rÛ  rÝ  rà  rä  ræ  rè  rê  rí  rð  ró  rö  rù  rû  rý  r½   rT   rZ   rX   ú<module>r     sÎ  ððð ð 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 Ø Ð Ð Ð Ð Ð Ð Ð Ð à rÐ rÐ rÐ rÐ rÐ rÐ rÐ rÐ rÐ rÐ rÐ rØ }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ð }Ø GÐ GÐ GÐ GÐ GÐ GÐ GÐ GØ /Ð /Ð /Ð /Ð /Ð /ðJð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jð Jà FÐ FÐ FÐ FÐ FÐ FÐ FÐ FÐ FÐ FØ 3Ð 3Ð 3Ð 3Ð 3Ð 3Ð 3Ð 3Ø YÐ YÐ YÐ YÐ YÐ YÐ YÐ YÐ YÐ YàÐ/Ð
0€ðð ð ð ð �"”)ñ ô ð ðD
}&ð }&ð }&ð }&ð }&˜"œ)ñ }&ô }&ð }&ð@!ð !˜#ð !¨4ð !Àkð !ð !ð !ð !ðJ KPð+ð +Øð+Ø*/ð+ØCGð+àð+ð +ð +ð +ð` %(Ø"%Ø$(Ø ðmð mØðmà!ðmð  ðmð ˜S”Mð	mð
 ðmð ðmð mð mð mð`Dð D˜#ð D°5ð DÈDð DÐgrð Dð Dð Dð DðN*ð *˜ð *°ð *È$ð *Ðepð *ð *ð *ð *ð^ %(Ø$(Ø ð	@ð @Øð@à!ð@ð ˜S”Mð@ð ð	@ð ð@ð @ð @ð @ðFð ˆcð  T¨#¨s¨(¤^ð ð ð ð ð& %Ð$ð W&Ø% t t° |¡|¤|ðW&à# T TØð EØð&ñ &ô &ðW&ð 0°°ØØÐ&=ÐCYØ }ÀCð2Iñ 2Iô 2IðW&ð /°°Ø  |¸Ø@Ø&ð OØð1ñ 1ô 1ðW&ð  '¨¨ð MØØ@Ø&Ø  |¸ÐSXð)Zñ )Zô )Zð!W&ð, 1°$°$ØØ$Ð*@Ø ¨4¸6ÐS`Ðpsð3uñ 3uô 3uð-W&ð6 & t tð HØØð(!ñ (!ô (!ð7W&ð> & t tð HØØð(!ñ (!ô (!ð?W&ðF & t tð GØØð(!ñ (!ô (!ðGW&ðP   Ø{ØØð!!ñ !!ô !!ðQW&ðZ $ T Tð EØØ$Ð*@ð&Bñ &Bô &Bð[W&ðb $ T Tð EØØ$Ð*@ð&Bñ &Bô &BðcW&ðj ,¨T¨Tð MØØ$Ð*@ð.Bñ .Bô .BðkW&ðr $ T Tð EØØ$Ð*@ð&Bñ &Bô &BðsW&ðz $ T Tð FØØ$Ð*@ð&Bñ &Bô &Bð{W&ðB ,¨T¨Tð MØØ$Ð*@ð.Bñ .Bô .BðCW&ðL ˜$˜$ØyØØ%°ð6ñ 6ô 6ðMW&ð W&ðT ˜$˜$ØyØØ%°ð6ñ 6ô 6ðUW&ð\ ˜$˜$ØyØØ °-È$ÐZ`ðbñ bô bð]W&ðf ˜4˜4™6œ6ðgW&ðh ˜$˜$ØuØØðñ ô ðiW&ðr ˜$˜$ØuØØðñ ô ðsW&ð| ˜$˜$ØuØØðñ ô ð}W&ðF ˜4˜4™6œ6ðGW&ðJ +¨D¨DØ$Ð*@Ø%°TÈð-Tñ -Tô -TðKW&ðP 1°$°$ØØ$Ð*@Ø%°TÈð3Tñ 3Tô 3TðQW&ðX -¨d¨dØØ%°TÈð/Tñ /Tô /TðYW&ð^ -¨d¨dØØ%°TÈð/Tñ /Tô /Tð_W&ðd -¨d¨dØØ$Ð*@Ø ¨F¸TØ%°SÈ	ð	/Sñ /Sô /SðeW&ðn -¨d¨dØØ ¨FØ }ÀCÐW`ð/bñ /bô /bðoW&ðv -¨d¨dØØ }ÀCÐW`ð/bñ /bô /bðwW&ð~ 6°t°tØØ ¨HØ Yð80ñ 80ô 80ðW&ðF .¨t¨tØØØ ¨HØ Ið	0/ñ 0/ô 0/ðGW&ðP 1°$°$ØØØ ¨HØ Ið	3/ñ 3/ô 3/ðQW&ð W&ð W&ðZ .¨t¨tØØØ ¨HØ Ið	0/ñ 0/ô 0/ð[W&ðf ,¨T¨TØØ ¨HØ }ÀCÐW`ð.bñ .bô .bðgW&ðn ,¨T¨TØØ ¨FØ }ÀCÐW`ð.bñ .bô .bðoW&ðx 8¸¸ØØ ¨FØ }ÀCÐW`ð:bñ :bô :bðyW&ð@ 2°4°4ØØ ¨FØ }ÀCÐW`ð4bñ 4bô 4bðAW&ðH 2°4°4ØØ ¨HØ }ÀCÐW`ð4bñ 4bô 4bðIW&ðP /°°ØØ }ÀCÐW`ð1bñ 1bô 1bðQW&ðV 0°°ØØØ ¨FØ }ÀCÐW`ð	2bñ 2bô 2bðWW&ð` 1°$°$ØØ ¨HØ }ÀCÐW`ð3bñ 3bô 3bðaW&ðh .¨t¨tØØ ¨HØ }ÀCÐW`ð0bñ 0bô 0bðiW&ðt +¨D¨Dà ¨F¸TÐQZð-\ñ -\ô -\ðuW&ðz 2°4°4ØØ }ÀCÐW`ð4bñ 4bô 4bð{W&ð@ 8¸¸ØØ ¨HØ }ÀCÐW`ð:bñ :bô :bðAW&ðH 8¸¸ØØ ¨HØ }ÀCÐW`ð:bñ :bô :bðIW&ðP /°°ØØ ¨HØ }ÀCÐW`ð1bñ 1bô 1bðQW&ðX 0°°ØØØ ¨HØ }ÀCÐW`ð	2bñ 2bô 2bðYW&ðb .¨t¨tà Yð00ñ 00ô 00ðcW&ð W&ðh -¨d¨dà ¨F¸TÐQZð/\ñ /\ô /\ðiW&ð W&ñ Wô W€ðt ðð  dð Àð ð ð ñ „ðð ðð  dð Àð ð ð ñ „ðð
 ðð  tð À+ð ð ð ñ „ðð
 ðð  dð Àð ð ð ñ „ðð ðð  dð Àð ð ð ñ „ðð ðð  dð Àð ð ð ñ „ðð ðð ˜tð ¸+ð ð ð ñ „ðð ðð ¨ð ÀKð ð ð ñ „ðð ðð ¨ð ÀKð ð ð ñ „ðð ðð °ð ÈKð ð ð ñ „ðð ðð ¨ð ÀKð ð ð ñ „ðð ðð ¨ð ÀKð ð ð ñ „ðð ðð °ð ÈKð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ˜$ð °[ð ð ð ñ „ðð ðð ¨4ð Àkð ð ð ñ „ðð ðð ¨4ð Àkð ð ð ñ „ðð ðð  tð À+ð ð ð ñ „ðð ðð ¨ð À;ð ð ð ñ „ðð ðð  tð À+ð ð ð ñ „ðð ðð ¨$ð À[ð ð ð ñ „ðð ðð ¨ð ÀKð ð ð ñ „ðð ðð ¨4ð Àkð ð ð ñ „ðð ðð ¨Tð Èð ð ð ñ „ðð ðð ¨$ð À[ð ð ð ñ „ðð ðð ¨dð Èð ð ð ñ „ðð ðð ¨Tð Èð ð ð ñ „ðð Ð ˜HØ"LØ(Jð'ð 'ñ ô ð ð ð rZ   