Ë
    P/æjH  ã                   óp   — d dl mZ d dlZd dlmZ d dlmZ d dlmZm	Z	m
Z
mZ d dlmZ dgZ G d„ de«      Zy)	é    )ÚNumberN)Úconstraints)ÚDistribution)Úbroadcast_allÚlazy_propertyÚlogits_to_probsÚprobs_to_logits)Ú binary_cross_entropy_with_logitsÚ	Geometricc                   óþ   ‡ — e Zd ZdZej
                  ej                  dœZej                  Z	dˆ fd„	Z
dˆ fd„	Zed„ «       Zed„ «       Zed„ «       Zed„ «       Zed	„ «       Z ej(                  «       fd
„Zd„ Zd„ Zˆ xZS )r   a€  
    Creates a Geometric distribution parameterized by :attr:`probs`,
    where :attr:`probs` is the probability of success of Bernoulli trials.

    .. math::

        P(X=k) = (1-p)^{k} p, k = 0, 1, ...

    .. note::
        :func:`torch.distributions.geometric.Geometric` :math:`(k+1)`-th trial is the first success
        hence draws samples in :math:`\{0, 1, \ldots\}`, whereas
        :func:`torch.Tensor.geometric_` `k`-th trial is the first success hence draws samples in :math:`\{1, 2, \ldots\}`.

    Example::

        >>> # xdoctest: +IGNORE_WANT("non-deterministic")
        >>> m = Geometric(torch.tensor([0.3]))
        >>> m.sample()  # underlying Bernoulli has 30% chance 1; 70% chance 0
        tensor([ 2.])

    Args:
        probs (Number, Tensor): the probability of sampling `1`. Must be in range (0, 1]
        logits (Number, Tensor): the log-odds of sampling `1`.
    )ÚprobsÚlogitsc           
      ó"  •— |d u |d u k(  rt        d«      ‚|�t        |«      \  | _        nt        |«      \  | _        |�|n|}t	        |t
        «      rt        j                  «       }n|j                  «       }t        ‰	| �)  ||¬«       | j                  r{|�x| j                  }|dkD  }|j                  «       sV|j                  |    }t        dt        |«      j                  › dt!        |j"                  «      › dt%        | «      › d|› �«      ‚y y y )Nz;Either `probs` or `logits` must be specified, but not both.©Úvalidate_argsr   zExpected parameter probs (z
 of shape z) of distribution z* to be positive but found invalid values:
)Ú
ValueErrorr   r   r   Ú
isinstancer   ÚtorchÚSizeÚsizeÚsuperÚ__init__Ú_validate_argsÚallÚdataÚtypeÚ__name__ÚtupleÚshapeÚrepr)
Úselfr   r   r   Úprobs_or_logitsÚbatch_shapeÚvalueÚvalidÚinvalid_valueÚ	__class__s
            €ú`/root/repos/pocs/video-evals/.venv/lib/python3.12/site-packages/torch/distributions/geometric.pyr   zGeometric.__init__/   s   ø€ Ø�TˆM˜v¨˜~Ò.ÜØMóð ð ÐÜ)¨%Ó0‰MˆT�Zä*¨6Ó2‰NˆTŒ[Ø#(Ð#4™%¸&ˆÜ�o¤vÔ.ÜŸ*™*›,‰Kà)×.Ñ.Ó0ˆKÜ‰Ñ˜°MÐÔBØ×Ò 5Ð#4à—J‘JˆEØ˜A‘IˆEØ—9‘9”;Ø %§
¡
¨E¨6Ñ 2�Ü ðÜ˜U›×,Ñ,Ð-¨Z¼¸e¿k¹kÓ8JÐ7Kð L'Ü'+¨D£z lð 3AØANÀðQóð ð ð	 $5Ðó    c                 ób  •— | j                  t        |«      }t        j                  |«      }d| j                  v r | j
                  j                  |«      |_        d| j                  v r | j                  j                  |«      |_        t        t        |�'  |d¬«       | j                  |_
        |S )Nr   r   Fr   )Ú_get_checked_instancer   r   r   Ú__dict__r   Úexpandr   r   r   r   )r!   r#   Ú	_instanceÚnewr'   s       €r(   r-   zGeometric.expandK   sŽ   ø€ Ø×(Ñ(¬°IÓ>ˆÜ—j‘j Ó-ˆØ�d—m‘mÑ#ØŸ
™
×)Ñ)¨+Ó6ˆCŒIØ�t—}‘}Ñ$ØŸ™×+Ñ+¨KÓ8ˆCŒJÜŒi˜Ñ& {À%Ð&ÔHØ!×0Ñ0ˆÔØˆ
r)   c                 ó&   — d| j                   z  dz
  S ©Ng      ð?©r   ©r!   s    r(   ÚmeanzGeometric.meanV   s   € à�T—Z‘ZÑ #Ñ%Ð%r)   c                 ó@   — t        j                  | j                  «      S ©N)r   Ú
zeros_liker   r3   s    r(   ÚmodezGeometric.modeZ   s   € ä×Ñ §
¡
Ó+Ð+r)   c                 ó@   — d| j                   z  dz
  | j                   z  S r1   r2   r3   s    r(   ÚvariancezGeometric.variance^   s   € à�d—j‘jÑ  3Ñ&¨$¯*©*Ñ4Ð4r)   c                 ó0   — t        | j                  d¬«      S ©NT)Ú	is_binary)r	   r   r3   s    r(   r   zGeometric.logitsb   s   € ä˜tŸz™z°TÔ:Ð:r)   c                 ó0   — t        | j                  d¬«      S r<   )r   r   r3   s    r(   r   zGeometric.probsf   s   € ä˜tŸ{™{°dÔ;Ð;r)   c                 óŠ  — | j                  |«      }t        j                  | j                  j                  «      j
                  }t        j                  «       5  t        j                  j                  «       rSt        j                  || j                  j                  | j                  j                  ¬«      }|j                  |¬«      }n+| j                  j                  |«      j                  |d«      }|j                  «       | j                   j                  «       z  j!                  «       cd d d «       S # 1 sw Y   y xY w)N)ÚdtypeÚdevice)Úminé   )Ú_extended_shaper   Úfinfor   r@   ÚtinyÚno_gradÚ_CÚ_get_tracing_stateÚrandrA   Úclampr/   Úuniform_ÚlogÚlog1pÚfloor)r!   Úsample_shaper   rF   Úus        r(   ÚsamplezGeometric.samplej   sÙ   € Ø×$Ñ$ \Ó2ˆÜ�{‰{˜4Ÿ:™:×+Ñ+Ó,×1Ñ1ˆÜ�]‰]‹_ñ 	=Ü�x‰x×*Ñ*Ô,ä—J‘J˜u¨D¯J©J×,<Ñ,<ÀTÇZÁZ×EVÑEVÔW�Ø—G‘G �GÓ%‘à—J‘J—N‘N 5Ó)×2Ñ2°4¸Ó;�Ø—E‘E“G §
¡
˜{×1Ñ1Ó3Ñ3×:Ñ:Ó<÷	=÷ 	=ò 	=ús   ÁCD9Ä9Ec                 ó(  — | j                   r| j                  |«       t        || j                  «      \  }}|j	                  t
        j                  ¬«      }d||dk(  |dk(  z  <   || j                  «       z  | j                  j                  «       z   S )N)Úmemory_formatr   rC   )	r   Ú_validate_sampler   r   Úcloner   Úcontiguous_formatrN   rM   )r!   r$   r   s      r(   Úlog_probzGeometric.log_probv   s~   € Ø×ÒØ×!Ñ! %Ô(Ü$ U¨D¯J©JÓ7‰ˆˆuØ—‘¬%×*AÑ*A�ÓBˆØ-.ˆˆu˜‰z˜e q™jÑ)Ñ*Ø˜˜—~‘~Ó'Ñ'¨$¯*©*¯.©.Ó*:Ñ:Ð:r)   c                 ó`   — t        | j                  | j                  d¬«      | j                  z  S )NÚnone)Ú	reduction)r
   r   r   r3   s    r(   ÚentropyzGeometric.entropy~   s(   € ä,¨T¯[©[¸$¿*¹*ÐPVÔWØ�j‰jñð	
r)   )NNNr6   )r   Ú
__module__Ú__qualname__Ú__doc__r   Úunit_intervalÚrealÚarg_constraintsÚnonnegative_integerÚsupportr   r-   Úpropertyr4   r8   r:   r   r   r   r   r   rR   rX   r\   Ú__classcell__)r'   s   @r(   r   r      sµ   ø„ ñð0 !,× 9Ñ 9À[×EUÑEUÑV€OØ×-Ñ-€Gõõ8	ð ñ&ó ð&ð ñ,ó ð,ð ñ5ó ð5ð ñ;ó ð;ð ñ<ó ð<ð #- %§*¡*£,ó 
=ò;ö
r)   )Únumbersr   r   Útorch.distributionsr   Ú torch.distributions.distributionr   Útorch.distributions.utilsr   r   r   r	   Útorch.nn.functionalr
   Ú__all__r   © r)   r(   ú<module>rn      s8   ðå ã Ý +Ý 9÷ó õ Að ˆ-€ôo
�õ o
r)   