§
    ·¯¸j  ã                   óº   — d Z ddlZddlmZ ddlmZmZ ddlZddlmZ ddl	m
Z ddlmZ dd	lmZ dd
lmZ ddlmZ d„ Ze G d„ dej        ¦  «        ¦   «         ZdS )zå PyTorch Conditionally Parameterized Convolution (CondConv)

Paper: CondConv: Conditionally Parameterized Convolutions for Efficient Inference
(https://arxiv.org/abs/1904.04971)

Hacked together by / Copyright 2020 Ross Wightman
é    N)Úpartial)ÚUnionÚTuple)Únn)Ú
functionalé   )Úregister_notrace_module)Ú	to_2tuple)Úconv2d_same)Úget_padding_valuec                 ó   ‡ ‡‡— ˆˆ ˆfd„}|S )Nc                 ó.  •— t          j        ‰¦  «        }t          | j        ¦  «        dk    s"| j        d         ‰k    s| j        d         |k    rt	          d¦  «        ‚t          ‰¦  «        D ]&} ‰| |                              ‰¦  «        ¦  «         Œ'dS )zCondConv initializer function.é   r   r   z<CondConv variables must have shape [num_experts, num_params]N)ÚmathÚprodÚlenÚshapeÚ
ValueErrorÚrangeÚview)ÚweightÚ
num_paramsÚiÚexpert_shapeÚinitializerÚnum_expertss      €€€úg/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/timm/layers/cond_conv2d.pyÚcondconv_initializerz6get_condconv_initializer.<locals>.condconv_initializer   s¤   ø€ å”Y˜|Ñ,Ô,ˆ
Ý�”ÑÔ Ò"Ð" f¤l°1¤o¸Ò&DÐ&DØ”˜Q” :Ò-Ð-ÝØNñPô Pð Qå�{Ñ#Ô#ð 	6ð 	6ˆAØˆK˜˜qœ	Ÿš |Ñ4Ô4Ñ5Ô5Ð5Ð5ð	6ð 	6ó    © )r   r   r   r   s   ``` r   Úget_condconv_initializerr!      s0   øøø€ ð6ð 6ð 6ð 6ð 6ð 6ð 6ð  Ðr   c                   óò   ‡ — e Zd ZdZg d¢Z	 	 	 	 	 	 	 	 	 dd	ed
edeeeeef         f         deeeeef         f         deeeeef         ef         deeeeef         f         dede	defˆ fd„Z
d„ Zd„ Zˆ xZS )Ú
CondConv2daO   Conditionally Parameterized Convolution
    Inspired by: https://github.com/tensorflow/tpu/blob/master/models/official/efficientnet/condconv/condconv_layers.py

    Grouped convolution hackery for parallel execution of the per-sample kernel filters inspired by this discussion:
    https://github.com/pytorch/pytorch/issues/17983
    )Úin_channelsÚout_channelsÚdynamic_paddingé   r   Ú Fé   Nr$   r%   Úkernel_sizeÚstrideÚpaddingÚdilationÚgroupsÚbiasr   c                 ó2  •— |
|dœ}t          ¦   «                              ¦   «          || _        || _        t	          |¦  «        | _        t	          |¦  «        | _        t          ||||¬¦  «        \  }}|| _        t	          |¦  «        | _	        t	          |¦  «        | _
        || _        |	| _        | j        | j        | j        z  f| j        z   | _        d}| j        D ]}||z  }Œt          j                             t          j        | j        |fi |¤Ž¦  «        | _        |rM| j        f| _        t          j                             t          j        | j        | j        fi |¤Ž¦  «        | _        n|                      dd ¦  «         |                      ¦   «          d S )N)ÚdeviceÚdtype)r+   r-   r   r/   )ÚsuperÚ__init__r$   r%   r
   r*   r+   r   r&   r,   r-   r.   r   Úweight_shapeÚtorchr   Ú	ParameterÚemptyr   Ú
bias_shaper/   Úregister_parameterÚreset_parameters)Úselfr$   r%   r*   r+   r,   r-   r.   r/   r   r1   r2   ÚddÚpadding_valÚis_padding_dynamicÚweight_num_paramÚwdÚ	__class__s                    €r   r4   zCondConv2d.__init__.   s¢  ø€ ð ¨Ð/Ð/ˆÝ‰Œ×ÒÑÔÐà&ˆÔØ(ˆÔÝ$ [Ñ1Ô1ˆÔÝ Ñ'Ô'ˆŒÝ*;Ø�[¨¸(ð+Dñ +Dô +DÑ'ˆÐ'à1ˆÔÝ  Ñ-Ô-ˆŒÝ! (Ñ+Ô+ˆŒØˆŒØ&ˆÔà!Ô.°Ô0@ÀDÄKÑ0OÐPÐSWÔScÑcˆÔØÐØÔ#ð 	#ð 	#ˆBØ Ñ"ÐÐÝ”h×(Ò(­¬°TÔ5EÐGWÐ)^Ð)^Ð[]Ð)^Ð)^Ñ_Ô_ˆŒàð 	2Ø#Ô0Ð2ˆDŒOÝœ×*Ò*­5¬;°tÔ7GÈÔIZÐ+aÐ+aÐ^`Ð+aÐ+aÑbÔbˆDŒIˆIà×#Ò# F¨DÑ1Ô1Ð1à×ÒÑÔÐÐÐr   c                 óÔ  — t          t          t          j        j        t          j        d¦  «        ¬¦  «        | j        | j        ¦  «        } || j	        ¦  «         | j
        �…t          j        | j        dd …         ¦  «        }dt          j        |¦  «        z  }t          t          t          j        j        | |¬¦  «        | j        | j        ¦  «        } || j
        ¦  «         d S d S )Né   )Úar   )rE   Úb)r!   r   r   ÚinitÚkaiming_uniform_r   Úsqrtr   r5   r   r/   r   Úuniform_r9   )r<   Úinit_weightÚfan_inÚboundÚ	init_biass        r   r;   zCondConv2d.reset_parametersY   sÙ   € Ý.Ý•B”GÔ,µ´	¸!±´Ð=Ñ=Ô=¸tÔ?OÐQUÔQbñdô dˆàˆ�D”KÑ Ô Ð ØŒ9Ð Ý”Y˜tÔ0°°°Ô4Ñ5Ô5ˆFØ�œ	 &Ñ)Ô)Ñ)ˆEÝ0Ý�œÔ(¨U¨F°eÐ<Ñ<Ô<¸dÔ>NÐPTÔP_ñaô aˆIàˆI�d”iÑ Ô Ð Ð Ð ð !Ð r   c           
      óÒ  — |j         \  }}}}t          j        || j        ¦  «        }|| j        z  | j        | j        z  f| j        z   }|                     |¦  «        }d }	| j	        �7t          j        || j	        ¦  «        }	|	                     || j        z  ¦  «        }	| 
                    d||z  ||¦  «        }| j        r.t          |||	| j        | j        | j        | j        |z  ¬¦  «        }
n2t!          j        |||	| j        | j        | j        | j        |z  ¬¦  «        }
|
                     g d¢¦  «                             || j        |
j         d         |
j         d         ¦  «        }
|
S )Nr   )r+   r,   r-   r.   )r   r   r   r'   éþÿÿÿéÿÿÿÿ)r   r6   Úmatmulr   r%   r$   r.   r*   r   r/   Úreshaper&   r   r+   r,   r-   ÚFÚconv2dÚpermute)r<   ÚxÚrouting_weightsÚBÚCÚHÚWr   Únew_weight_shaper/   Úouts              r   ÚforwardzCondConv2d.forwardd   so  € Ø”W‰
ˆˆ1ˆa�Ý”˜o¨t¬{Ñ;Ô;ˆØ Ô 1Ñ1°4Ô3CÀtÄ{Ñ3RÐSÐVZÔVfÑfÐØ—’Ð-Ñ.Ô.ˆØˆØŒ9Ð Ý”< °´Ñ;Ô;ˆDØ—9’9˜Q Ô!2Ñ2Ñ3Ô3ˆDð �IŠI�a˜˜Q™  1Ñ%Ô%ˆØÔð 	@ÝØ�6˜4¨¬¸T¼\Øœ¨t¬{¸Q©ð@ñ @ô @ˆCˆCõ ”(Ø�6˜4¨¬¸T¼\Øœ¨t¬{¸Q©ð@ñ @ô @ˆCð �kŠk˜,˜,˜,Ñ'Ô'×,Ò,¨Q°Ô0AÀ3Ä9ÈRÄ=ÐRUÔR[Ð\^ÔR_Ñ`Ô`ˆð& ˆ
r   )	r'   r   r(   r   r   Fr)   NN)Ú__name__Ú
__module__Ú__qualname__Ú__doc__Ú__constants__Úintr   r   ÚstrÚboolr4   r;   r_   Ú__classcell__)rB   s   @r   r#   r#   $   s8  ø€ € € € € ðð ð GÐFÐF€Mð 89Ø23Ø8:Ø45ØØØ ØØð) ð ) àð) ð ð) ð ˜s E¨#¨s¨(¤OÐ3Ô4ð	) ð
 ˜#˜u S¨# XœÐ.Ô/ð) ð ˜3  c¨3 h¤°Ð4Ô5ð) ð ˜C  s¨C x¤Ð0Ô1ð) ð ð) ð ð) ð ð) ð ) ð ) ð ) ð ) ð ) ðV	!ð 	!ð 	!ð'ð 'ð 'ð 'ð 'ð 'ð 'r   r#   )rc   r   Ú	functoolsr   Útypingr   r   r6   r   Útorch.nnr   rT   Ú_fxr	   Úhelpersr
   r   r,   r   r!   ÚModuler#   r    r   r   ú<module>ro      s  ððð ð €€€Ø Ð Ð Ð Ð Ð Ø Ð Ð Ð Ð Ð Ð Ð à €€€Ø Ð Ð Ð Ð Ð Ø $Ð $Ð $Ð $Ð $Ð $à (Ð (Ð (Ð (Ð (Ð (Ø Ð Ð Ð Ð Ð Ø $Ð $Ð $Ð $Ð $Ð $Ø &Ð &Ð &Ð &Ð &Ð &ð
 ð 
 ð 
 ð ðfð fð fð fð f�”ñ fô fñ Ôðfð fð fr   