o
    h{                     @   sj  U d dl Z d dlZd dlmZ d dlmZmZ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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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. ddl/m0Z0m1Z1 ddl2m3Z3 ddl4m5Z5 ddl6m7Z7 ddl8m9Z9 ddl:m;Z; ddl<m=Z= ddl>m?Z?m@ZA i ZBi ZCeeeef ef eDd< d d!gZEd"d  ZFeG d#d$ d$ZGd%d& ZHd'd( ZId)d* ZJd+d, ZKd-ed.ed/e	jLfd0d!ZMeFeed1d2 ZNeFeed3d4 ZOeFeed5d6 ZPeFeed7d8 ZQeFeed9d: ZReFeed;d< ZSeFeed=d> ZTeFeed?d@ ZUeFe e dAdB ZVeFe$e$dCdD ZWeFe"e"dEdF ZXeFe&e&dGdH ZYeFe*e*dIdJ ZZeFe.e.dKdL Z[eFe1e.dMdN Z\eFe.e1dOdP Z]eFe1e1dQdR Z^eFe3e3dSdT Z_eFe5e5dUdV Z`eFe7e7dWdX ZaeFe9e9dYdZ ZbeFe;e;d[d\ ZceFe=e=d]d^ ZdeFee9d_d` ZeeFeedadb ZfeFee7dcdd ZgeFeededf ZheFee dgdh ZieFee3didj ZjeFee=dkdl ZkeFee7dmdn ZleFeedodp ZmeFee3dqdr ZneFee=dsdt ZoeFeeeFeeeFee7eFee=dudv ZpeFee dwdx ZqeFee$dydz ZreFee3d{d| ZseFe eeFe eeFe e7eFe e=d}d~ ZteFe edd ZueFe e$dd ZveFe e3dd ZweFe$eeFe$eeFe$eeFe$e eFe$e7eFe$e=dd ZxeFe$e3dd ZyeFe*eeFe*eeFe*eeFe*e eFe*e7eFe*e=dd ZzeFe*e3dd Z{eFe3eeFe3eeFe3eeFe3e eFe3e7eFe3e=dd Z|eFe3e$dd Z}eFe3e*dd Z~eFe7eeFe7eeFe7e=dd ZeFe7edd ZeFe7e dd ZeFe7e3dd ZeFe9eeFe9edd ZeFe=edd ZeFe=edd ZeFe=edd ZeFe=e dd ZeFe=e$dd ZeFe=e3dd ZeFe=e7dd ZeFe(e(dd ZeFeedd Zdd ZdS )    N)total_ordering)CallableDictTupleType)inf   )	Bernoulli)Beta)Binomial)Categorical)Cauchy)ContinuousBernoulli)	Dirichlet)Distribution)ExponentialFamily)Exponential)Gamma)	Geometric)Gumbel)
HalfNormal)Independent)Laplace)_batch_lowrank_logdet_batch_lowrank_mahalanobisLowRankMultivariateNormal)_batch_mahalanobisMultivariateNormal)Normal)OneHotCategorical)Pareto)Poisson)TransformedDistribution)Uniform)_sum_rightmosteuler_constant_KL_MEMOIZEregister_klkl_divergencec                    sV   t  tst trtd  t ts"ttr"td  fdd}|S )a[  
    Decorator to register a pairwise function with :meth:`kl_divergence`.
    Usage::

        @register_kl(Normal, Normal)
        def kl_normal_normal(p, q):
            # insert implementation here

    Lookup returns the most specific (type,type) match ordered by subclass. If
    the match is ambiguous, a `RuntimeWarning` is raised. For example to
    resolve the ambiguous situation::

        @register_kl(BaseP, DerivedQ)
        def kl_version1(p, q): ...
        @register_kl(DerivedP, BaseQ)
        def kl_version2(p, q): ...

    you should register a third most-specific implementation, e.g.::

        register_kl(DerivedP, DerivedQ)(kl_version1)  # Break the tie.

    Args:
        type_p (type): A subclass of :class:`~torch.distributions.Distribution`.
        type_q (type): A subclass of :class:`~torch.distributions.Distribution`.
    z6Expected type_p to be a Distribution subclass but got z6Expected type_q to be a Distribution subclass but got c                    s   | t  f< t  | S N)_KL_REGISTRYr&   clear)funtype_ptype_q L/var/www/html/ai/venv/lib/python3.10/site-packages/torch/distributions/kl.py	decoratorT   s   zregister_kl.<locals>.decorator)
isinstancetype
issubclassr   	TypeError)r.   r/   r2   r0   r-   r1   r'   1   s   c                   @   s*   e Zd ZdgZdd Zdd Zdd ZdS )	_Matchtypesc                 G   s
   || _ d S r)   r8   )selfr8   r0   r0   r1   __init__`   s   
z_Match.__init__c                 C   s   | j |j kS r)   r9   )r:   otherr0   r0   r1   __eq__c   s   z_Match.__eq__c                 C   s:   t | j|jD ]\}}t||s dS ||ur dS qdS )NFT)zipr8   r5   )r:   r<   xyr0   r0   r1   __le__f   s   
z_Match.__le__N)__name__
__module____qualname__	__slots__r;   r=   rA   r0   r0   r0   r1   r7   \   s
    r7   c           	         s    fddt D }|stS tdd |D j\}}tdd |D j\}}t ||f }t ||f }||urFtd jj|j|jt |S )zP
    Find the most specific approximate match, assuming single inheritance.
    c                    s,   g | ]\}}t  |rt |r||fqS r0   )r5   ).0super_psuper_qr-   r0   r1   
<listcomp>s   s    z _dispatch_kl.<locals>.<listcomp>c                 s   s    | ]}t | V  qd S r)   )r7   rF   mr0   r0   r1   	<genexpr>}   s    z_dispatch_kl.<locals>.<genexpr>c                 s   s    | ]	}t t| V  qd S r)   )r7   reversedrJ   r0   r0   r1   rL   ~   s    z;Ambiguous kl_divergence({}, {}). Please register_kl({}, {}))	r*   NotImplementedminr8   warningswarnformatrB   RuntimeWarning)	r.   r/   matchesleft_pleft_qright_qright_pleft_fun	right_funr0   r-   r1   _dispatch_klo   s"   r[   c                 C   s   t | tS )zI
    Helper function for obtaining infinite KL Divergence throughout
    )torch	full_liker   tensorr0   r0   r1   _infinite_like      r`   c                 C   s   | |    S )z2
    Utility function for calculating x log x
    )logr^   r0   r0   r1   _x_log_x   ra   rc   c                 C   sD   |  d}|  d}| d|| dd}|| jdd S )zp
    Utility function for calculating the trace of XX^{T} with X having arbitrary trailing batch dimensions
       N)sizereshapepowsumshape)bmatnrK   
flat_tracer0   r0   r1   _batch_trace_XXT   s   

ro   pqreturnc                 C   s   zt t| t|f }W n ty(   tt| t|}|t t| t|f< Y nw |tu r;td| jj d|jj || |S )a"  
    Compute Kullback-Leibler divergence :math:`KL(p \| q)` between two distributions.

    .. math::

        KL(p \| q) = \int p(x) \log\frac {p(x)} {q(x)} \,dx

    Args:
        p (Distribution): A :class:`~torch.distributions.Distribution` object.
        q (Distribution): A :class:`~torch.distributions.Distribution` object.

    Returns:
        Tensor: A batch of KL divergences of shape `batch_shape`.

    Raises:
        NotImplementedError: If the distribution types have not been registered via
            :meth:`register_kl`.
    z(No KL(p || q) is implemented for p type z and q type )r&   r4   KeyErrorr[   rN   NotImplementedError	__class__rB   )rp   rq   r,   r0   r0   r1   r(      s   
c                 C   s   | j tjj|j tjj| j   }t||j dk< d|| j dk< d| j  tjj|jtjj| j  }t||j dk< d|| j dk< || S Nr   r   )probsr\   nn
functionalsoftpluslogitsr   rp   rq   t1t2r0   r0   r1   _kl_bernoulli_bernoulli   s   r   c           	      C   s   | j | j }|j |j }|j  |j  |  }| j  | j  |  }| j |j  t| j  }| j|j t| j }|| t| }|| | | | S r)   )concentration1concentration0lgammar\   digamma)	rp   rq   sum_params_psum_params_qr}   r~   t3t4t5r0   r0   r1   _kl_beta_beta   s   r   c                 C   sh   | j |j k  rtd| j | j| j|j  | j   |j    }| j |j k}t|| ||< |S )NzKKL between Binomials where q.total_count > p.total_count is not implemented)total_countanyrt   rw   r{   log1pr`   )rp   rq   klinf_idxsr0   r0   r1   _kl_binomial_binomial   s   (r   c                 C   sD   | j | j|j  }t||j dk|< d|| j dk|< |dS )Nr   rd   )rw   r{   r   	expand_asrj   )rp   rq   tr0   r0   r1   _kl_categorical_categorical   s   
r   c                 C   sL   | j | j|j  }|  t| j  }|  t|j  }|| | S r)   )meanr{   _cont_bern_log_normr\   r   rw   rp   rq   r}   r~   r   r0   r0   r1   -_kl_continuous_bernoulli_continuous_bernoulli   s   r   c                 C   s|   | j d}|j d}| |  }| j  |j   d}| j |j  }| j  | d }|| || d S )Nrd   )concentrationrj   r   r   	unsqueeze)rp   rq   sum_p_concentrationsum_q_concentrationr}   r~   r   r   r0   r0   r1   _kl_dirichlet_dirichlet  s   r   c                 C   s"   |j | j  }|  }|| d S Nr   raterb   )rp   rq   
rate_ratior}   r0   r0   r1   _kl_exponential_exponential  s   
r   c                 C   s   t | t |kstddd | jD }|j}| j| }tjj| |dd}|j| | }t|||D ]\}}}	|| |	 }
|t	|
t
|j8 }q4|S )NzThe cross KL-divergence between different exponential families cannot                             be computed using Bregman divergencesc                 S   s   g | ]}|   qS r0   )detachrequires_grad_)rF   npr0   r0   r1   rI     s    z+_kl_expfamily_expfamily.<locals>.<listcomp>T)create_graph)r4   rt   _natural_params_log_normalizerr\   autogradgradrj   r>   r$   lenevent_shape)rp   rq   	p_nparams	q_nparams	lg_normal	gradientsresultpnpqnpgtermr0   r0   r1   _kl_expfamily_expfamily  s   
r   c                 C   sn   |j | j|j   }t|j t| j  }| j |j  t| j  }|j| j | j | j  }|| | | S r)   )r   r   rb   r\   r   r   rp   rq   r}   r~   r   r   r0   r0   r1   _kl_gamma_gamma*  s
   r   c                 C   sl   | j |j  }|j|j  }| j|j  }|  | | }|t }t|d|   | }|| | dt  S r   )scalelocrb   _euler_gammar\   expr   )rp   rq   ct1ct2ct3r}   r~   r   r0   r0   r1   _kl_gumbel_gumbel3  s   r   c                 C   s$   |    t|j | j  |j S r)   )entropyr\   r   rw   r{   rp   rq   r0   r0   r1   _kl_geometric_geometric>  s   $r   c                 C      t | j|jS r)   )_kl_normal_normal	base_distr   r0   r0   r1   _kl_halfnormal_halfnormalC     r   c                 C   sV   | j |j  }| j|j  }|  }||j  }|t| | j   }|| | d S r   )r   r   absrb   r\   r   )rp   rq   scale_ratioloc_abs_diffr}   r~   r   r0   r0   r1   _kl_laplace_laplaceH  s   

r   c                 C   s   | j |j kr
tdt|j|j|jt| j| j| j }t|j|j|j| j |j}|jj|j	d }t
jj|j|dd}| j|j d}t| j|j 	d }t|| j 	d }t|| j}	|| | |	 }
d||
 | | j d   S )NzKL-divergence between two Low Rank Multivariate Normals with                          different event shapes cannot be computedre   Fupperrd         ?r   )r   
ValueErrorr   _unbroadcasted_cov_factor_unbroadcasted_cov_diag_capacitance_trilr   r   mTr   r\   linalgsolve_triangularrj   ro   rsqrtsqrtmatmul)rp   rq   term1term3	qWt_qDinvAterm21term22term23term24term2r0   r0   r1   7_kl_lowrankmultivariatenormal_lowrankmultivariatenormalS  s6   
	r   c           	      C   s   | j |j kr
tdt|j|j|jd| jjddd 	d  }t
|j|j|j| j |j}|jj|jd }tjj|j|dd}t| j|j d }t|| j}|| }d|| | | j d	   S )
NKL-divergence between two (Low Rank) Multivariate Normals with                          different event shapes cannot be computedrf   re   rd   dim1dim2Fr   r   r   )r   r   r   r   r   r   _unbroadcasted_scale_trildiagonalrb   rj   r   r   r   r   r\   r   r   ro   r   r   )	rp   rq   r   r   r   r   r   r   r   r0   r0   r1   0_kl_multivariatenormal_lowrankmultivariatenormalu  s.   
	r   c                 C   s$  | j |j kr
tdd|jjddd d t| j| j| j	 }t
|j|j| j }tj|jjd d | jjd d }| j d }|j|||f }| j||| jdf }t| j |||f }ttjj||dd}	ttjj||dd}
|	|
 }d	|| | | j d   S )
Nr   rf   re   rd   r   r   Fr   r   )r   r   r   r   rb   rj   r   r   r   r   r   r   r\   _C_infer_sizerk   expand
cov_factorrg   
diag_embedr   ro   r   r   )rp   rq   r   r   combined_batch_shaperm   q_scale_trilp_cov_factor
p_cov_diagr   r   r   r0   r0   r1   0_kl_lowrankmultivariatenormal_multivariatenormal  s>   

r   c           	      C   s   | j |j kr
td|jjddd d| jjddd d }tj|jj	d d | jj	d d }| j d }|j
|||f }| j
|||f }ttjj||dd}t|j|j| j }|d|| |   S )	NzvKL-divergence between two Multivariate Normals with                          different event shapes cannot be computedre   rd   r   r   Fr   r   )r   r   r   r   rb   rj   r\   r   r   rk   r   ro   r   r   r   r   )	rp   rq   
half_term1r   rm   r   p_scale_trilr   r   r0   r0   r1   )_kl_multivariatenormal_multivariatenormal  s(   
r   c                 C   sB   | j |j  d}| j|j |j  d}d|| d |   S Nrf   r   r   r   ri   r   rb   )rp   rq   	var_ratior}   r0   r0   r1   r     s   r   c                 C   r   r)   )r   _categoricalr   r0   r0   r1   '_kl_onehotcategorical_onehotcategorical  r   r   c                 C   sX   | j |j  }|j| j }|j|  }|  }|| | d }t|| jj|jjk < |S r   )r   alpharb   r   supportlower_bound)rp   rq   r   alpha_ratior}   r~   r   r0   r0   r1   _kl_pareto_pareto  s   
r   c                 C   s&   | j | j  |j    | j |j   S r)   r   r   r0   r0   r1   _kl_poisson_poisson     &r   c                 C   s.   | j |j krt| j|jkrtt| j|jS r)   )
transformsrt   r   r(   r   r   r0   r0   r1   _kl_transformed_transformed  s
   r  c                 C   s<   |j |j | j | j   }t||j| jk|j | j k B < |S r)   )highlowrb   r   rp   rq   r   r0   r0   r1   _kl_uniform_uniform  s   r  c                 C   s    |    | j|j  |j  S r)   )r   rw   r   rb   r   r0   r0   r1   _kl_bernoulli_poisson  s    r  c                 C   s,   |    | j|j  t|j  |  S r)   )r   r   r{   r\   r   rw   r   r   r0   r0   r1   _kl_beta_continuous_bernoulli  s   
r	  c                 C   
   t | jS r)   )r`   r   r   r0   r0   r1   _kl_beta_infinity     
r  c                 C   s,   |    |j  |j| j| j| j    S r)   )r   r   rb   r   r   r   r0   r0   r1   _kl_beta_exponential  s   r  c                 C   sp   |    }|j |j|j   }|jd | j | j| j    }|j| j | j| j  }|| | | S r   )r   r   r   r   rb   r   r   r   r   r0   r0   r1   _kl_beta_gamma  s   
r  c           	      C   s   | j | j | j  }|jd}|   }d|d tj   }|d|  | j | j d  |d d }|j| }|jdd }|| || | |  S r   )	r   r   r   ri   r   mathpirb   r   )	rp   rq   E_beta
var_normalr}   r~   r   r   r   r0   r0   r1   _kl_beta_normal*  s   

r  c                 C   s>   |    |j|j   }t||j| jjk|j| jjk B < |S r)   )r   r  r  rb   r   r   r   upper_boundr  r0   r0   r1   _kl_beta_uniform9  s    r  c                 C   r
  r)   )r`   rw   r   r0   r0   r1   !_kl_continuous_bernoulli_infinityC  r  r  c                 C   s"   |    t|j |j| j  S r)   )r   r\   rb   r   r   r   r0   r0   r1   $_kl_continuous_bernoulli_exponentialH  s   "r  c                 C   sz   |    }dtdtj t|j|j   t|j }| jt| j	 d|j | j	  dt|j  }|| | S )Nr   g       @)
r   r  rb   r  r\   squarer   r   variancer   r   r0   r0   r1   _kl_continuous_bernoulli_normalQ  s   
( r  c              	   C   sV   |    |j|j   }ttt|j| jj	t
|j| jjt|t |S r)   )r   r  r  rb   r\   wheremaxger   r   ler  	ones_liker   r  r0   r0   r1    _kl_continuous_bernoulli_uniform]  s   r   c                 C   r
  r)   r`   r   r   r0   r0   r1   _kl_exponential_infinityj     
r"  c                 C   sB   |j | j  }|j t| }|| |j  |jt  dt  S r   )r   r   r\   rb   r   r   )rp   rq   ratior}   r0   r0   r1   _kl_exponential_gammar  s   r%  c                 C   sR   | j |j }|j|j }| d }t|| |d  }| }|| | | S r   )r   r   r   rb   r\   r   
reciprocal)rp   rq   scale_rate_prodloc_scale_ratior}   r~   r   r0   r0   r1   _kl_exponential_gumbel  s   r)  c                 C   sp   |j d}| jd}dt|| d tj  }| }|j| j }|jdd }|d || | |  S r   )	r   ri   r   r\   rb   r  r  r&  r   )rp   rq   r  rate_sqrr}   r~   r   r   r0   r0   r1   _kl_exponential_normal  s   r+  c                 C   r
  r)   )r`   r   r   r0   r0   r1   _kl_gamma_infinity  r#  r,  c                 C   s&   |    |j  |j| j | j  S r)   )r   r   rb   r   r   r0   r0   r1   _kl_gamma_exponential  r  r-  c                 C   s~   | j |j }|j|j }| jd | j  | j  | j }| | j|  }t|d|	  
| j  | }|| | S r   )r   r   r   r   r   r   rb   r\   r   r&  ri   )rp   rq   beta_scale_prodr(  r}   r~   r   r0   r0   r1   _kl_gamma_gumbel  s    r/  c                 C   s   |j d}| jd}dt|| d tj  | j | j  }d| jd| j  | }|j	| j | j }d|j	d }|| jd | j
   || | |  S r   )r   ri   r   r\   rb   r  r  r   r   r   r   )rp   rq   r  beta_sqrr}   r~   r   r   r0   r0   r1   _kl_gamma_normal  s"   r1  c                 C   r
  r)   r`   r   r   r0   r0   r1   _kl_gumbel_infinity     
r3  c                 C   sx   | j |j  }|tdtj   }tj| d dd }| j| j t  |j |j  dd }| | | td  S )Nrf   r      r   )r   r  r   r  rb   ri   r   r   )rp   rq   param_ratior}   r~   r   r0   r0   r1   _kl_gumbel_normal  s
   &r7  c                 C   r
  r)   r2  r   r0   r0   r1   _kl_laplace_infinity  r4  r8  c                 C   s~   |j d}| j d| }dtd| tj  }d| jd }| j|j }d|jd }| | || | |  d S r   )r   ri   r\   rb   r  r  r   )rp   rq   r  scale_sqr_var_ratior}   r~   r   r   r0   r0   r1   _kl_laplace_normal  s   r:  c                 C   r
  r)   r2  r   r0   r0   r1   _kl_normal_infinity  r4  r;  c                 C   s|   | j |j }| j|j d}|j |j }| d }|| }t| d|  | }| | | ddtdtj    S r   )r   r   ri   rb   r\   r   r  r  )rp   rq   mean_scale_ratiovar_scale_sqr_ratior(  r}   r~   r   r0   r0   r1   _kl_normal_gumbel  s   &r>  c                 C   s   | j |j  }| j|j }|| j }t|}tdtj | j td|d  }|t	td|  }| || |j  ddtdtj    S )Nrf   g      r   r   )
r   r   r\   rb   r  r   r  r   ri   erf)rp   rq   loc_diffr   loc_diff_scale_ratior}   r~   r   r0   r0   r1   _kl_normal_laplace  s   

(,rB  c                 C   r
  r)   )r`   r   r   r0   r0   r1   _kl_pareto_infinity  s   
rC  c                 C   sZ   | j |j }| j|  }| j }| j| | jd  }|| | d }t|| jdk< |S r   )r   r   r   rb   r&  r   )rp   rq   r'  r}   r~   r   r   r0   r0   r1   _kl_pareto_exponential   s   
rD  c                 C   s   | j  | j  }| j | }|j |j|j   }d|j | }|j| j | j  | jd  }|| | | d }t|| jdk< |S r   )r   rb   r   r&  r   r   r   r   rp   rq   common_termr}   r~   r   r   r   r0   r0   r1   _kl_pareto_gamma+  s   rG  c           	      C   s   d|j d }| j | jd  }tdtj |j  | j | j   }| j }| j|d | jd  }| j| |j d}|| || |  d }t	|| jdk< |S )Nrf   r   )
r   ri   r   r  r   r  rb   r&  r   r   )	rp   rq   r  rF  r}   r~   r   r   r   r0   r0   r1   _kl_pareto_normal:  s   &
rH  c                 C   r
  r)   r!  r   r0   r0   r1   _kl_poisson_infinityG  s   
rI  c                 C   s   | j | j }t|}|jd t| j t| j |  | }|jd td| j  td| j  |  | }|j |j  |j|j   }|| | | }t|| j |j	j
k| j|j	jk B < |S r   )r  r  r\   rb   r   rc   r   r   r   r   r  r   rE  r0   r0   r1   _kl_uniform_betaM  s.   
 rJ  c              	   C   sh   |    | j|j  t|j  |  }ttt	| j
|jjt| j|jjt|t |S r)   )r   r   r{   r\   r   rw   r   r  r  r  r  r   r  r  r  r   r  r   r  r0   r0   r1    _kl_uniform_continuous_bernoullie  s    
rK  c                 C   sB   |j | j| j  d | j| j |j    }t|| j|jjk < |S )Nrf   )r   r  r  rb   r   r   r   r  r0   r0   r1   _kl_uniform_exponetialw  s   ,rL  c                 C   s   | j | j }| }|j |j|j   }d|j t| j t| j |  | }|j| j | j  d }| | | | }t|| j|jj	k < |S )Nr   rf   )
r  r  rb   r   r   r   rc   r   r   r   rE  r0   r0   r1   _kl_uniform_gamma~  s   rM  c                 C   sn   |j | j| j  }| j|j |j  }| j|j |j  }| d||   }|t| t|   }|| S )Nr   )r   r  r  r   rb   r\   r   )rp   rq   rF  high_loc_difflow_loc_diffr}   r~   r0   r0   r1   _kl_uniform_gumbel  s   rP  c                 C   st   | j | j }ttjd |j |  }|dd }| j | j d|j  d d}|d||  |jd  S )Nrf      r   )	r  r  r  r   r  r   rb   ri   r   )rp   rq   rF  r}   r~   r   r0   r0   r1   _kl_uniform_normal  s
    rR  c                 C   sl   | j | j }|j|j|j |  }t| j t| j | | }||jd  | }t|| j|jj	k < |S r   )
r  r  r   r   ri   rb   rc   r   r   r   )rp   rq   support_uniformr}   r~   r   r0   r0   r1   _kl_uniform_pareto  s   rT  c                 C   s*   | j |j krtt| j|j}t|| j S r)   )reinterpreted_batch_ndimsrt   r(   r   r$   r  r0   r0   r1   _kl_independent_independent  s   rV  c                 C   sD   | j |j  d| j|j d  }d| j  |j   }|| S )Nrf      r   r|   r0   r0   r1   _kl_cauchy_cauchy  s   (rX  c                  C   sb   dg} t tdd dD ]\}}| d|j d|j d qd| }tjr/t j|7  _d	S d	S )
zHAppends a list of implemented KL functions to the doc for kl_divergence.zLKL divergence is currently implemented for the following distribution pairs:c                 S   s   | d j | d j fS rv   )rB   )p_qr0   r0   r1   <lambda>  s    z_add_kl_info.<locals>.<lambda>)keyz* :class:`~torch.distributions.z#` and :class:`~torch.distributions.`z
	N)sortedr*   appendrB   joinr(   __doc__)rowsrp   rq   kl_infor0   r0   r1   _add_kl_info  s   
rc  )r  rP   	functoolsr   typingr   r   r   r   r\   r   	bernoullir	   betar
   binomialr   categoricalr   cauchyr   continuous_bernoullir   	dirichletr   distributionr   
exp_familyr   exponentialr   gammar   	geometricr   gumbelr   half_normalr   independentr   laplacer   lowrank_multivariate_normalr   r   r   multivariate_normalr   r   normalr   one_hot_categoricalr   paretor    poissonr!   transformed_distributionr"   uniformr#   utilsr$   r%   r   r*   r&   __annotations____all__r'   r7   r[   r`   rc   ro   Tensorr(   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r  r  r  r	  r  r  r  r  r  r  r  r  r   r"  r%  r)  r+  r,  r-  r/  r1  r3  r7  r8  r:  r;  r>  rB  rC  rD  rG  rH  rI  rJ  rK  rL  rM  rP  rR  rT  rV  rX  rc  r0   r0   r0   r1   <module>   s  
 +
&















!

$








	




	



























	

