
    /j                         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mZmZmZmZ d dlmZmZmZ d	d
gZ G d d	e          Z G d d
e          ZdS )    N)Tensor)constraints)Distribution)TransformedDistribution)SigmoidTransform)broadcast_allclamp_probslazy_propertylogits_to_probsprobs_to_logits)_Number_sizeNumberLogitRelaxedBernoulliRelaxedBernoullic                   .    e Zd ZdZej        ej        dZej        Z	 	 	 dde	de	e
z  dz  de	e
z  dz  dedz  ddf
 fd	Zd fd
	Zd Zede	fd            Zede	fd            Zedej        fd            Z ej                    fdede	fdZd Z xZS )r   a  
    Creates a LogitRelaxedBernoulli distribution parameterized by :attr:`probs`
    or :attr:`logits` (but not both), which is the logit of a RelaxedBernoulli
    distribution.

    Samples are logits of values in (0, 1). See [1] for more details.

    Args:
        temperature (Tensor): relaxation temperature
        probs (Number, Tensor): the probability of sampling `1`
        logits (Number, Tensor): the log-odds of sampling `1`

    [1] The Concrete Distribution: A Continuous Relaxation of Discrete Random
    Variables (Maddison et al., 2017)

    [2] Categorical Reparametrization with Gumbel-Softmax
    (Jang et al., 2017)
    probslogitsNtemperaturer   r   validate_argsreturnc                    || _         |d u |d u k    rt          d          |,t          |t                    }t	          |          \  | _        n<|t          d          t          |t                    }t	          |          \  | _        || j        n| j        | _        |rt          j
                    }n| j                                        }t                                          ||           d S )Nz;Either `probs` or `logits` must be specified, but not both.zlogits is unexpectedly Noner   )r   
ValueError
isinstancer   r   r   AssertionErrorr   _paramtorchSizesizesuper__init__)selfr   r   r   r   	is_scalarbatch_shape	__class__s          j/home/longshao/multi-rider-rag/.venv/lib/python3.11/site-packages/torch/distributions/relaxed_bernoulli.pyr#   zLogitRelaxedBernoulli.__init__.   s     'TMv~..M   "5'22I)%00MTZZ~$%BCCC"6733I*622NT[$)$5djj4; 	-*,,KK+**,,KMBBBBB    c                    |                      t          |          }t          j        |          }| j        |_        d| j        v r+| j                            |          |_        |j        |_        d| j        v r+| j	                            |          |_	        |j	        |_        t          t          |                              |d           | j        |_        |S )Nr   r   Fr   )_get_checked_instancer   r   r    r   __dict__r   expandr   r   r"   r#   _validate_argsr$   r&   	_instancenewr'   s       r(   r-   zLogitRelaxedBernoulli.expandK   s    (()>	JJj--*dm##
))+66CICJt}$$++K88CJCJ#S))22;e2TTT!0
r)   c                 &     | j         j        |i |S N)r   r1   )r$   argskwargss      r(   _newzLogitRelaxedBernoulli._newY   s    t{////r)   c                 .    t          | j        d          S NT)	is_binary)r   r   r$   s    r(   r   zLogitRelaxedBernoulli.logits\   s    tzT::::r)   c                 .    t          | j        d          S r8   )r   r   r:   s    r(   r   zLogitRelaxedBernoulli.probs`   s    t{d;;;;r)   c                 4    | j                                         S r3   )r   r!   r:   s    r(   param_shapez!LogitRelaxedBernoulli.param_shaped   s    {!!!r)   sample_shapec                    |                      |          }t          | j                            |                    }t          t	          j        ||j        |j                            }|                                | 	                                z
  |                                z   | 	                                z
  | j
        z  S )N)dtypedevice)_extended_shaper	   r   r-   r   randr@   rA   loglog1pr   )r$   r>   shaper   uniformss        r(   rsamplezLogitRelaxedBernoulli.rsampleh   s    $$\22DJ--e4455JuEKEEE
 
 LLNNxi..000599;;>5&AQAQQ 	r)   c                 0   | j         r|                     |           t          | j        |          \  }}||                    | j                  z
  }| j                                        |z   d|                                                                z  z
  S )N   )	r.   _validate_sampler   r   mulr   rD   exprE   )r$   valuer   diffs       r(   log_probzLogitRelaxedBernoulli.log_probr   s     	)!!%(((%dk599		$"2333##%%,q488::3C3C3E3E/EEEr)   NNNr3   )__name__
__module____qualname____doc__r   unit_intervalrealarg_constraintssupportr   r   boolr#   r-   r6   r
   r   r   propertyr   r    r=   r   rH   rP   __classcell__r'   s   @r(   r   r      s        ( !, 9[EUVVOG
 )-)-%)C CC %C $&	C
 d{C 
C C C C C C:     0 0 0 ; ; ; ; ]; <v < < < ]< "UZ " " " X" -7EJLL  E V    F F F F F F Fr)   c                        e Zd ZU dZej        ej        dZej        ZdZ	e
ed<   	 	 	 ddedeez  dz  deez  dz  d	edz  d
df
 fdZd fd	Zed
efd            Zed
efd            Zed
efd            Z xZS )r   a  
    Creates a RelaxedBernoulli distribution, parametrized by
    :attr:`temperature`, and either :attr:`probs` or :attr:`logits`
    (but not both). This is a relaxed version of the `Bernoulli` distribution,
    so the values are in (0, 1), and has reparametrizable samples.

    Example::

        >>> # xdoctest: +IGNORE_WANT("non-deterministic")
        >>> m = RelaxedBernoulli(torch.tensor([2.2]),
        ...                      torch.tensor([0.1, 0.2, 0.3, 0.99]))
        >>> m.sample()
        tensor([ 0.2951,  0.3442,  0.8918,  0.9021])

    Args:
        temperature (Tensor): relaxation temperature
        probs (Number, Tensor): the probability of sampling `1`
        logits (Number, Tensor): the log-odds of sampling `1`
    r   T	base_distNr   r   r   r   r   c                     t          |||          }t                                          |t                      |           d S )Nr   )r   r"   r#   r   )r$   r   r   r   r   r_   r'   s         r(   r#   zRelaxedBernoulli.__init__   sB     *+ufEE	$4$6$6mTTTTTr)   c                     |                      t          |          }t                                          ||          S )N)r0   )r+   r   r"   r-   r/   s       r(   r-   zRelaxedBernoulli.expand   s3    (()99EEww~~kS~999r)   c                     | j         j        S r3   )r_   r   r:   s    r(   r   zRelaxedBernoulli.temperature   s    ~))r)   c                     | j         j        S r3   )r_   r   r:   s    r(   r   zRelaxedBernoulli.logits   s    ~$$r)   c                     | j         j        S r3   )r_   r   r:   s    r(   r   zRelaxedBernoulli.probs   s    ~##r)   rQ   r3   )rR   rS   rT   rU   r   rV   rW   rX   rY   has_rsampler   __annotations__r   r   rZ   r#   r-   r[   r   r   r   r\   r]   s   @r(   r   r   z   sl         ( !, 9[EUVVO'GK$$$$
 )-)-%)U UU %U $&	U
 d{U 
U U U U U U: : : : : : *V * * * X* % % % % X% $v $ $ $ X$ $ $ $ $r)   )r   r   torch.distributionsr    torch.distributions.distributionr   ,torch.distributions.transformed_distributionr   torch.distributions.transformsr   torch.distributions.utilsr   r	   r
   r   r   torch.typesr   r   r   __all__r   r    r)   r(   <module>ro      sM          + + + + + + 9 9 9 9 9 9 P P P P P P ; ; ; ; ; ;              / . . . . . . . . . #$6
7aF aF aF aF aFL aF aF aFH4$ 4$ 4$ 4$ 4$. 4$ 4$ 4$ 4$ 4$r)   