
    iE                         d Z ddlmZ ddlZddlmZ ddlmZmZmZm	Z	m
Z
mZmZmZmZmZmZmZmZmZ ddgZ G d	 de      Zd
de de de
 de de dz   e_         dee   dee   dee   dee   dee   dededededededededdfdZdee   dee   dee   dee   dee   dededededededededdfdZ e	e      	 	 	 	 	 d"dee   dee   dee   dee   dee   d edz  dededededededededdfd!       Zy)#z1Implementation for the Resilient backpropagation.    )castN)Tensor   )_capturable_doc_default_to_fused_or_foreach_differentiable_doc_disable_dynamo_if_unsupported_foreach_doc!_get_capturable_supported_devices_get_scalar_dtype_maximize_doc_params_doc
_to_scalar_use_grad_for_differentiable_view_as_real	OptimizerParamsTRproprpropc                        e Zd Z	 	 	 dddddddedeez  deeef   deeef   ded	edz  d
ededdf fdZ fdZ	d Z
edd       Z xZS )r   FN)
capturableforeachmaximizedifferentiableparamslretas
step_sizesr   r   r   r   returnc                   t        |t              r|j                         dk7  rt        d      d|k  st        d|       d|d   cxk  rdcxk  r|d   k  sn t        d|d    d|d          |||||||d	}	t        
|   ||	       y )
Nr   zTensor lr must be 1-elementg        zInvalid learning rate: r         ?zInvalid eta values: z, )r   r   r   r   r   r   r   )
isinstancer   numel
ValueErrorsuper__init__)selfr   r   r   r   r   r   r   r   defaults	__class__s             K/var/www/html/engine/venv/lib/python3.12/site-packages/torch/optim/rprop.pyr&   zRprop.__init__   s     b&!bhhjAo:;;by6rd;<<T!W,s,T!W,3DG9BtAwiHII $ ,$
 	*    c                 0   t         |   |       | j                  D ]  }|j                  dd        |j                  dd       |j                  dd       |j                  dd       |d   D ]  }| j                  j                  |g       }t        |      dk7  s.t        j                  |d         rGt        |d         }|d   r*t        j                  |t               |j                  	      nt        j                  |t               
      |d<     y )Nr   r   Fr   r   r   r   stepdtypedevicer/   )r%   __setstate__param_groups
setdefaultstategetlentorch	is_tensorfloattensorr   r0   )r'   r5   grouppp_statestep_valr)   s         r*   r2   zRprop.__setstate__=   s    U#&& 	EY-Z/-u5\518_ 
**..B/w<1$U__WV_-M$WV_5H
 !. $,=,? #\\(:K:MN FO	
	r+   c           	      d   d}|d   D ]  }|j                   |t        j                  |      z  }|j                  |       |j                   }	|	j                  rt        d      |j                  |	       | j                  |   }
t        |
      dk(  r|d   r*t        j                  dt               |j                        nt        j                  dt                     |
d	<   t        j                  |t        j                  
      |
d<   |j                  j                  r*t        j                  |	t        |d   |d               |
d<   n%t        j                  |	t!        |d               |
d<   |j                  |
d          |j                  |
d          |j                  |
d	           |S )NFr   z'Rprop does not support sparse gradientsr   r    r.   r1   r-   memory_formatprevr   	step_size)gradr8   
is_complexappend	is_sparseRuntimeErrorr5   r7   zerosr   r0   
zeros_likepreserve_formatr/   	full_likecomplexr   )r'   r<   r   gradsprevsr   state_stepshas_complexr=   rF   r5   s              r*   _init_groupzRprop._init_groupP   s|   x  	.Avv~5++A..KMM!66D~~"#LMMLLJJqME 5zQ \* KK*;*=ahhOR/@/BC f !& 0 0%BWBW Xf77%% */geDk5;?*E+& */z%PT+?V)WE+&LLv'eK01uV}-A 	.D r+   c                 b   | j                          d}|$t        j                         5   |       }ddd       | j                  D ][  }g }g }g }g }g }|d   \  }	}
|d   \  }}|d   }|d   }| j	                  ||||||      }t        ||||||||	|
|||d   |d   |       ] |S # 1 sw Y   uxY w)	zPerform a single optimization step.

        Args:
            closure (Callable, optional): A closure that reevaluates the model
                and returns the loss.
        Nr   r   r   r   r   r   )	step_size_minstep_size_maxetaminusetaplusr   r   r   r   rS   ) _cuda_graph_capture_health_checkr8   enable_gradr3   rT   r   )r'   closurelossr<   r   rP   rQ   r   rR   rX   rY   rV   rW   r   r   rS   s                   r*   r-   z
Rprop.stepv   s	    	--/""$ !y! && 	E#%F"$E"$E')J(*K %fHg+0+>(M=I&GZ(H**vueZK ++!!$%56 .'!	B I! !s   B%%B.)g{Gz?)g      ?g333333?)gư>2   N)__name__
__module____qualname__r   r:   r   tupleboolr&   r2   rT   r   r-   __classcell__)r)   s   @r*   r   r      s     "$.*4+ !#$++ FN+ E5L!	+
 %,'+ + + + + 
+<&$L "/ "/r+   a
  Implements the resilient backpropagation algorithm.

    .. math::
       \begin{aligned}
            &\rule{110mm}{0.4pt}                                                                 \\
            &\textbf{input}      : \theta_0 \in \mathbf{R}^d \text{ (params)},f(\theta)
                \text{ (objective)},                                                             \\
            &\hspace{13mm}      \eta_{+/-} \text{ (etaplus, etaminus)}, \Gamma_{max/min}
                \text{ (step sizes)}                                                             \\
            &\textbf{initialize} :   g^0_{prev} \leftarrow 0,
                \: \eta_0 \leftarrow \text{lr (learning rate)}                                   \\
            &\rule{110mm}{0.4pt}                                                                 \\
            &\textbf{for} \: t=1 \: \textbf{to} \: \ldots \: \textbf{do}                         \\
            &\hspace{5mm}g_t           \leftarrow   \nabla_{\theta} f_t (\theta_{t-1})           \\
            &\hspace{5mm} \textbf{for} \text{  } i = 0, 1, \ldots, d-1 \: \mathbf{do}            \\
            &\hspace{10mm}  \textbf{if} \:   g^i_{prev} g^i_t  > 0                               \\
            &\hspace{15mm}  \eta^i_t \leftarrow \mathrm{min}(\eta^i_{t-1} \eta_{+},
                \Gamma_{max})                                                                    \\
            &\hspace{10mm}  \textbf{else if}  \:  g^i_{prev} g^i_t < 0                           \\
            &\hspace{15mm}  \eta^i_t \leftarrow \mathrm{max}(\eta^i_{t-1} \eta_{-},
                \Gamma_{min})                                                                    \\
            &\hspace{15mm}  g^i_t \leftarrow 0                                                   \\
            &\hspace{10mm}  \textbf{else}  \:                                                    \\
            &\hspace{15mm}  \eta^i_t \leftarrow \eta^i_{t-1}                                     \\
            &\hspace{5mm}\theta_t \leftarrow \theta_{t-1}- \eta_t \mathrm{sign}(g_t)             \\
            &\hspace{5mm}g_{prev} \leftarrow  g_t                                                \\
            &\rule{110mm}{0.4pt}                                                          \\[-1.ex]
            &\bf{return} \:  \theta_t                                                     \\[-1.ex]
            &\rule{110mm}{0.4pt}                                                          \\[-1.ex]
       \end{aligned}

    For further details regarding the algorithm we refer to the paper
    `A Direct Adaptive Method for Faster Backpropagation Learning: The RPROP Algorithm
    <http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.21.1417>`_.z

    Args:
        a{  
        lr (float, optional): learning rate (default: 1e-2)
        etas (Tuple[float, float], optional): pair of (etaminus, etaplus), that
            are multiplicative increase and decrease factors
            (default: (0.5, 1.2))
        step_sizes (Tuple[float, float], optional): a pair of minimal and
            maximal allowed step sizes (default: (1e-6, 50))
        z	
        z

    r   rP   rQ   r   rR   rV   rW   rX   rY   r   r   r   rS   r   c                   t        |       D ]  \  }}||   }|	s|n| }||   }||   }||   }t        j                  j                         s`|
r^t	               }|j
                  j                  |j
                  j                  k(  r|j
                  j                  |v st        d| d      |dz  }t        j                  |      rTt        j                  |      }t        j                  |      }t        j                  |      }t        j                  |      }|r.|j                  |j                               j                         }n|j                  |      j                         }|
r|j                  t        j                  |j                  d      ||             |j                  t        j                  |j!                  d      ||             |j                  t        j                  |j#                  d      d|             n<|||j                  d      <   |||j!                  d      <   d||j#                  d      <   |j%                  |      j'                  ||       |j                  t        j(                        }|
r6|j                  t        j                  |j#                  |      d|             nd||j#                  |      <   |j+                  |j                         |d       |j                  |        y )NIIf capturable=True, params and state_steps must be on supported devices: .r   r   rB   value)	enumerater8   compileris_compilingr   r0   typeAssertionErrorrG   view_as_realmulclonesigncopy_wheregtlteqmul_clamp_rM   addcmul_)r   rP   rQ   r   rR   rV   rW   rX   rY   r   r   r   rS   iparamrF   rD   rE   r-   capturable_supported_devicesrt   s                        r*   _single_tensor_rpropr      sp     f% 45Qx#t$QxqM	1~ ~~**,+L+N(!!T[[%5%55LL%%)EE$_`|_}}~  		E"%%d+D%%d+D&&u-E**95I88DJJL)..0D88D>&&(DJJu{{4771:w=>JJu{{4771:x>?JJu{{4771:q$78&D'D D 	t##M=A zz(=(=z>JJu{{4778#4a>?&'D"# 	tyy{IR8

4i4r+   c          
         t        |       dk(  ry |rt        d      t        j                  j	                         s;|
r9t               t        fdt        | |d      D              st        d d      t        j                  | ||||g      }|j                         D ]%  \  \  }}}}}}t        t        t           |      }t        t        t           |      }t        t        t           |      }t        t        t           |      }t        t        t           |      }t        j                  j	                         s=|d   j                  r.t        j                  |t        j                   dd	
      d       nt        j                  |d       |rt#        ||||       t        j$                  ||      }|	rt        j&                  |       t        j(                  ||       |	rt        j&                  |       |}t        j*                  |       |
r|D ]  }|j-                  t        j.                  |j1                  d      ||             |j-                  t        j.                  |j3                  d      ||             |j-                  t        j.                  |j5                  d      d|              nC|D ]>  }|||j1                  d      <   |||j3                  d      <   d||j5                  d      <   @ t        j6                  ||       |D ]  }|j9                  ||        t        |      }t;        t        |            D ]@  }||   j-                  t        j.                  ||   j5                  |      d||                B ~|D cg c]  }|j=                          }}t        j>                  |||d       ( y c c}w )Nr   z#_foreach ops don't support autogradc              3      K   | ]N  \  }}|j                   j                  |j                   j                  k(  xr |j                   j                  v  P y wr_   )r0   ro   ).0r=   r-   r   s      r*   	<genexpr>z&_multi_tensor_rprop.<locals>.<genexpr>?  sQ      
 4 HHMMT[[--- >!==>
s   AAT)strictrg   rh   r!   cpu)r0   )alphar   ri   rj   ) r7   rp   r8   rm   rn   r   allzipr   "_group_tensors_by_device_and_dtypevaluesr   listr   is_cpu_foreach_add_r;   r   _foreach_mul_foreach_neg__foreach_copy__foreach_sign_ru   rv   rw   rx   ry   _foreach_mul_r{   rangert   _foreach_addcmul_) r   rP   rQ   r   rR   rV   rW   rX   rY   r   r   r   rS   grouped_tensorsgrouped_params_grouped_grads_grouped_prevs_grouped_step_sizes_grouped_state_steps__grouped_paramsgrouped_gradsgrouped_prevsgrouped_step_sizesgrouped_state_stepssignsrt   rE   r}   rF   
grad_signsr   s                                   @r*   _multi_tensor_rpropr   &  st     6{aBCC >>&&(Z'H'J$ 
 v{4@
 

 ![\x[yyz{   BB	z;7O ""$J
 		 	d6lO<T&\>:T&\>:!$v,0CD"4<1EF ~~**,1DQ1G1N1N#U\\#e%DC  3Q7 }>P ""=-@&
 	]M:.%U# =

5;;twwqz7DAB

5;;twwqz8TBC

5;;twwqz1d;<=
  %#*TWWQZ #+TWWQZ #$TWWQZ % 	.6+ 	;I]M:	;
 ]+s=)* 	A!""E!HKK11mA6FG	  /<<ddiik<
<J(:"	
QJ
N =s   O)single_tensor_fnr   c
                z   t         j                  j                         st        d |D              st	        d      |t        | |d      \  }}|r)t         j                  j                         rt	        d      |r%t         j                  j                         st        }nt        } || |||||
|||||||	       y)zpFunctional API that performs rprop algorithm computation.

    See :class:`~torch.optim.Rprop` for details.
    c              3   P   K   | ]  }t        |t        j                           y wr_   )r"   r8   r   )r   ts     r*   r   zrprop.<locals>.<genexpr>  s       5()
1ell#5s   $&zPAPI has changed, `state_steps` argument must contain a list of singleton tensorsNF)	use_fusedz6torch.jit.script not supported with foreach optimizers)rV   rW   rX   rY   r   r   r   rS   )
r8   rm   rn   r   rJ   r   jitis_scriptingr   r   )r   rP   rQ   r   rR   r   r   r   r   rS   rV   rW   rX   rY   r   funcs                   r*   r   r     s    4 >>&&( 5-85 2 ^
 	
 1Ne

7 599))+STTuyy--/"###%r+   )NFFFF)__doc__typingr   r8   r   	optimizerr   r   r   r	   r
   r   r   r   r   r   r   r   r   r   __all__r   r   r:   rd   r   r   r   rA   r+   r*   <module>r      s   8      $ G
HI HX!LD	 
 	 
 		 		 		 E1 lDLD<D <D V	D
 fD D D D D D D D D 
DNo
Lo
<o
 <o
 V	o

 fo
 o
 o
 o
 o
 o
 o
 o
 o
 
o
l  1EF   ;L;<; <; V	;
 f; D[; ; ; ; ; ; ;  !;" #;$ 
%; G;r+   