§
    ­¸j÷  ã                   ó   — d Z ddlZddlmc mZ ddlmZmZ ddej        dej        deedf         d	e	ej                 fd
„Z
	 	 	 	 ddede	e         dz  dedz  dededz  d	efd„Zddej        ded	ej        fd„Z G d„ dej        ¦  «        ZdS )zJMathematical building blocks: MLP, inverse_sigmoid, accuracy, interpolate.é    N)ÚTensorÚnn©é   ÚoutputÚtargetÚtopk.Úreturnc                 óæ  — t          j        ¦   «         5  |                     ¦   «         dk    r(t          j        g | j        ¬¦  «        gcddd¦  «         S t          |¦  «        }|                     d¦  «        }|                      |ddd¦  «        \  }}|                     ¦   «         }| 	                    | 
                    dd¦  «                             |¦  «        ¦  «        }g }|D ]o}	|d|	…                              d¦  «                             ¦   «                              d¦  «        }
|                     |
                     d|z  ¦  «        ¦  «         Œp|cddd¦  «         S # 1 swxY w Y   dS )z7Computes the precision@k for the specified values of k.r   )ÚdeviceNr   Téÿÿÿÿg      Y@)ÚtorchÚno_gradÚnumelÚzerosr   ÚmaxÚsizer	   ÚtÚeqÚviewÚ	expand_asÚreshapeÚfloatÚsumÚappendÚmul_)r   r   r	   ÚmaxkÚ
batch_sizeÚ_ÚpredÚcorrectÚresÚkÚ	correct_ks              úb/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/rfdetr/models/math.pyÚaccuracyr&      s©  € õ 
Œ‰Œð ð Ø�<Š<‰>Œ>˜QÒÐÝ”K ¨6¬=Ð9Ñ9Ô9Ð:ðð ð ð ñ ô ð ð õ �4‰yŒyˆØ—[’[ ‘^”^ˆ
à—+’+˜d A t¨TÑ2Ô2‰ˆˆ4Ø�vŠv‰xŒxˆØ—'’'˜&Ÿ+š+ a¨Ñ,Ô,×6Ò6°tÑ<Ô<Ñ=Ô=ˆàˆØð 	;ð 	;ˆAØ   œ×+Ò+¨BÑ/Ô/×5Ò5Ñ7Ô7×;Ò;¸AÑ>Ô>ˆIØ�JŠJ�y—~’~ e¨jÑ&8Ñ9Ô9Ñ:Ô:Ð:Ð:Øðð ð ð ñ ô ð ð ð ð ð ð øøøð ð ð ð ð ð s   ”4E&ÁDE&Å&E*Å-E*ÚnearestÚinputr   Úscale_factorÚmodeÚalign_cornersc                 ó6   — t          j        | ||||¦  «        }|S )a  Equivalent to nn.functional.interpolate.

    Historically this wrapped ``torchvision`` to support empty batch sizes on old releases; modern ``torch`` (>=2.2)
    handles empty batches natively, so this now delegates directly to :func:`torch.nn.functional.interpolate`.
    )ÚFÚinterpolate)r(   r   r)   r*   r+   Úinterpolateds         r%   r.   r.   &   s!   € õ œ=¨°°lÀDÈ-ÑXÔX€LØÐó    çñhãˆµøä>ÚxÚepsc                 ó¼   — |                       dd¬¦  «        } |                       |¬¦  «        }d| z
                        |¬¦  «        }t          j        ||z  ¦  «        S )aÐ  Compute element-wise logit (inverse sigmoid) of a tensor.

    Clamps input to ``[0, 1]``, then returns ``log(x / (1 - x))`` with
    numerical stability guarded by ``eps``.

    Args:
        x: Input tensor with values in the range ``[0, 1]``.
        eps: Minimum clamp value applied to ``x`` and ``1 - x`` before the
            log operation. Guards against ``log(0)``.

    Returns:
        Tensor of the same shape as ``x`` containing the logit values.
    r   r   )Úminr   )r5   )Úclampr   Úlog)r2   r3   Úx1Úx2s       r%   Úinverse_sigmoidr:   6   sW   € ð 	
�Š�A˜1ˆÑÔ€AØ	
�Š�SˆÑ	Ô	€BØ
ˆa‰%�Š˜3ˆÑ	Ô	€BÝŒ9�R˜"‘WÑÔÐr0   c            
       óH   ‡ — e Zd ZdZdededededdf
ˆ fd„Zd	edefd
„Zˆ xZS )ÚMLPz4Very simple multi-layer perceptron (also called FFN)Ú	input_dimÚ
hidden_dimÚ
output_dimÚ
num_layersr
   Nc                 óÚ   •— t          ¦   «                              ¦   «          || _        |g|dz
  z  }t          j        d„ t          |g|z   ||gz   ¦  «        D ¦   «         ¦  «        | _        d S )Nr   c              3   óF   K  — | ]\  }}t          j        ||¦  «        V — Œd S )N)r   ÚLinear)Ú.0Únr#   s      r%   ú	<genexpr>zMLP.__init__.<locals>.<genexpr>Q   s0   è è € Ð#gÐ#g¹¸¸1¥B¤I¨a°¡O¤OÐ#gÐ#gÐ#gÐ#gÐ#gÐ#gr0   )ÚsuperÚ__init__r@   r   Ú
ModuleListÚzipÚlayers)Úselfr=   r>   r?   r@   ÚhÚ	__class__s         €r%   rH   zMLP.__init__M   so   ø€ Ý‰Œ×ÒÑÔÐØ$ˆŒØˆL˜J¨™NÑ+ˆÝ”mÐ#gÐ#gÅÀYÀKÐRSÁOÐUVÐZdÐYeÑUeÑ@fÔ@fÐ#gÑ#gÔ#gÑgÔgˆŒˆˆr0   r2   c                 ó¦   — t          | j        ¦  «        D ];\  }}|| j        dz
  k     rt          j         ||¦  «        ¦  «        n
 ||¦  «        }Œ<|S )zÑApply the MLP: ReLU on all layers except the last.

        Args:
            x: Input tensor of shape ``(..., input_dim)``.

        Returns:
            Output tensor of shape ``(..., output_dim)``.
        r   )Ú	enumeraterK   r@   r-   Úrelu)rL   r2   ÚiÚlayers       r%   ÚforwardzMLP.forwardS   s_   € õ " $¤+Ñ.Ô.ð 	Jð 	J‰HˆAˆuØ$%¨¬¸!Ñ(;Ò$;Ð$;•”�u�u˜Q‘x”xÑ Ô Ð ÀÀÀqÁÄˆAˆAØˆr0   )	Ú__name__Ú
__module__Ú__qualname__Ú__doc__ÚintrH   r   rT   Ú__classcell__)rN   s   @r%   r<   r<   J   s•   ø€ € € € € Ø>Ð>ðh #ð h°3ð hÀCð hÐUXð hÐ]að hð hð hð hð hð hð˜ð  Fð ð ð ð ð ð ð ð r0   r<   )r   )NNr'   N)r1   )rX   r   Útorch.nn.functionalr   Ú
functionalr-   r   ÚtuplerY   Úlistr&   r   ÚstrÚboolr.   r:   ÚModuler<   © r0   r%   ú<module>rc      sm  ðð QÐ Pà €€€Ø Ð Ð Ð Ð Ð Ð Ð Ð Ø Ð Ð Ð Ð Ð Ð Ð ðð �U”\ð ¨5¬<ð ¸uÀSÈ#ÀX¼ð ÐZ^Ð_dÔ_kÔZlð ð ð ð ð. "Ø!%ØØ!%ðð Øðà
ˆsŒ)�dÑ
ðð ˜$‘,ðð ð	ð
 ˜$‘;ðð ðð ð ð ð ð �u”|ð ¨%ð ¸5¼<ð ð ð ð ð(ð ð ð ð ˆ"Œ)ñ ô ð ð ð r0   