ó
    "EñiE  ã                    ó  • S r SSKJr  SSKrSSKJr  SSKJrJrJrJ	r	J
r
JrJrJrJrJrJrJrJrJr  SS/r " S	 S\5      rS
S\ S\ S\
 S\ S\ S3-   \l         S\\   S\\   S\\   S\\   S\\   S\S\S\S\S\S\S\S\SS4S jrS\\   S\\   S\\   S\\   S\\   S\S\S\S\S\S\S\S\SS4S jr\	" \S9     S"S\\   S\\   S\\   S\\   S\\   S \S-  S\S\S\S\S\S\S\S\SS4S! jj5       rg)#z1Implementation for the Resilient backpropagation.é    )ÚcastN)ÚTensoré   )Ú_capturable_docÚ_default_to_fused_or_foreachÚ_differentiable_docÚ_disable_dynamo_if_unsupportedÚ_foreach_docÚ!_get_capturable_supported_devicesÚ_get_scalar_dtypeÚ_maximize_docÚ_params_docÚ
_to_scalarÚ_use_grad_for_differentiableÚ_view_as_realÚ	OptimizerÚParamsTÚRpropÚrpropc                   ó®   ^ • \ rS rSr   SSSSSS.S\S\\-  S\\\4   S\\\4   S	\S
\S-  S\S\SS4U 4S jjjjr	U 4S jr
S r\SS j5       rSrU =r$ )r   é   FN)Ú
capturableÚforeachÚmaximizeÚdifferentiableÚparamsÚlrÚetasÚ
step_sizesr   r   r   r   Úreturnc                ó.  >• [        U[        5      (       a  UR                  5       S:w  a  [        S5      eSU::  d  [        SU 35      eSUS   s=:  a  Ss=:  a	  US   :  d  O  [        SUS    SUS    35      eUUUUUUUS	.n	[        T
U ]  X5        g )
Nr   zTensor lr must be 1-elementg        zInvalid learning rate: r   ç      ð?zInvalid eta values: z, )r   r   r   r   r   r   r   )Ú
isinstancer   ÚnumelÚ
ValueErrorÚsuperÚ__init__)Úselfr   r   r   r   r   r   r   r   ÚdefaultsÚ	__class__s             €ÚN/home/mande/repo/quber/.venv/lib/python3.13/site-packages/torch/optim/rprop.pyr'   ÚRprop.__init__   s¬   ø€ ô �bœ&×!Ñ! b§h¡h£j°A£oÜÐ:Ó;Ð;Ø�b‹yÜÐ6°r°dÐ;Ó<Ð<Ø�T˜!‘WÕ,˜sÕ, T¨!¡WÕ,ÜÐ3°D¸±G°9¸B¸tÀA¹w¸iÐHÓIÐIð ØØ$ØØ Ø,Ø$ñ
ˆô 	‰Ñ˜Õ*ó    c                 óP  >• [         TU ]  U5        U R                   GH  nUR                  SS 5        UR                  SS5        UR                  SS5        UR                  SS5        US    H°  nU R                  R                  U/ 5      n[        U5      S:w  d  M0  [        R                  " US   5      (       a  MP  [        US   5      nUS   (       a(  [        R                  " U[        5       UR                  S	9O[        R                  " U[        5       S
9US'   M²     GM     g )Nr   r   Fr   r   r   r   Ústep©ÚdtypeÚdevice©r1   )r&   Ú__setstate__Úparam_groupsÚ
setdefaultÚstateÚgetÚlenÚtorchÚ	is_tensorÚfloatÚtensorr   r2   )r(   r7   ÚgroupÚpÚp_stateÚstep_valr*   s         €r+   r4   ÚRprop.__setstate__=   só   ø€ Ü‰Ñ˜UÔ#Ø×&Õ&ˆEØ×Ñ˜Y¨Ô-Ø×Ñ˜Z¨Ô/Ø×ÑÐ-¨uÔ5Ø×Ñ˜\¨5Ô1Ø˜8”_�ØŸ*™*Ÿ.™.¨¨BÓ/�Ü�w“< 1Õ$¬U¯_ª_¸WÀV¹_×-MÓ-MÜ$ W¨V¡_Ó5�Hð
 ! ×.ô ŸšØ$Ô,=Ó,?ÈÏÉòô #Ÿ\š\¨(Ô:KÓ:MÑNð ˜F“Oô	 %ò 'r-   c           	      ó†  • SnUS    GH´  nUR                   c  M  U[        R                  " U5      -  nUR                  U5        UR                   n	U	R                  (       a  [        S5      eUR                  U	5        U R                  U   n
[        U
5      S:X  aä  US   (       a(  [        R                  " S[        5       UR                  S9O[        R                  " S[        5       S9U
S	'   [        R                  " U[        R                  S
9U
S'   UR                  R                  (       a+  [        R                  " U	[        US   US   5      5      U
S'   O&[        R                  " U	[!        US   5      5      U
S'   UR                  U
S   5        UR                  U
S   5        UR                  U
S	   5        GM·     U$ )NFr   z'Rprop does not support sparse gradientsr   r   © r0   r3   r/   ©Úmemory_formatÚprevr   Ú	step_size)Úgradr:   Ú
is_complexÚappendÚ	is_sparseÚRuntimeErrorr7   r9   Úzerosr   r2   Ú
zeros_likeÚpreserve_formatr1   Ú	full_likeÚcomplexr   )r(   r>   r   ÚgradsÚprevsr   Ústate_stepsÚhas_complexr?   rI   r7   s              r+   Ú_init_groupÚRprop._init_groupP   sw  € ØˆØ�x•ˆAØ�v‰v‰~ÙØœ5×+Ò+¨AÓ.Ñ.ˆKØ�M‰M˜!ÔØ—6‘6ˆDØ�~�~Ü"Ð#LÓMÐMà�L‰L˜ÔØ—J‘J˜q‘MˆEô �5‹z˜Q‹ð ˜\×*ô —K’K Ô*;Ó*=ÀaÇhÁhÒOäŸš RÔ/@Ó/BÑCð �f‘ô !&× 0Ò 0°Ä%×BWÑBWÑ X��f‘Ø—7‘7×%×%ô */¯ªØœg e¨D¡k°5¸±;Ó?ó*�E˜+Ò&ô */¯ª¸¼zÈ%ÐPTÉ+Ó?VÓ)W�E˜+Ñ&à�L‰L˜˜v™Ô'Ø×Ñ˜e KÑ0Ô1Ø×Ñ˜u V™}×-ñA !ðD Ðr-   c                 ód  • U R                  5         SnUb%  [        R                  " 5          U" 5       nSSS5        U R                   HT  n/ n/ n/ n/ n/ nUS   u  pšUS   u  p¼US   nUS   nU R	                  X4XVXx5      n[        UUUUUUUU	U
UUUS   US   US9  MV     U$ ! , (       d  f       Nt= f)	z�Perform a single optimization step.

Args:
    closure (Callable, optional): A closure that reevaluates the model
        and returns the loss.
Nr   r   r   r   r   r   )	Ústep_size_minÚstep_size_maxÚetaminusÚetaplusr   r   r   r   rV   )Ú'_accelerator_graph_capture_health_checkr:   Úenable_gradr5   rW   r   )r(   ÚclosureÚlossr>   r   rS   rT   r   rU   r\   r]   rZ   r[   r   r   rV   s                   r+   r/   Ú
Rprop.stepv   só   € ð 	×4Ñ4Ô6àˆØÑÜ×"Ò"Õ$Ù“y�÷ %ð ×&Ô&ˆEØ#%ˆFØ"$ˆEØ"$ˆEØ')ˆJØ(*ˆKà % f¡ÑˆHØ+0°Ñ+>Ñ(ˆMØ˜IÑ&ˆGØ˜ZÑ(ˆHà×*Ñ*Ø˜u¨ZóˆKô ØØØØØØ+Ø+Ø!ØØØ!Ø$Ð%5Ñ6Ø  Ñ.Ø'ôñ! 'ðB ˆ÷I %Õ$ús   «B!Â!
B/rD   )g{®Gáz„?)g      à?g333333ó?)g�íµ ÷Æ°>é2   ©N)Ú__name__Ú
__module__Ú__qualname__Ú__firstlineno__r   r<   r   ÚtupleÚboolr'   r4   rW   r   r/   Ú__static_attributes__Ú__classcell__)r*   s   @r+   r   r      s¼   ø† ð "Ø$.Ø*4ð+ð !Ø#ØØ$ò+àð+ð �F‰Nð+ð �E˜5�LÑ!ð	+ð
 ˜% ˜,Ñ'ð+ð ð+ð ˜‘ð+ð ð+ð ð+ð 
÷+ñ +õ<ò&$ðL "ó/ó "ö/r-   a¼
  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   rS   rT   r   rU   rZ   r[   r\   r]   r   r   r   rV   r    c                ó`  • [        U 5       GH  u  pÞX   nU	(       d  UOU* nX-   nX=   nXM   n[        R                  R                  5       (       dh  U
(       aa  [	        5       nUR
                  R                  UR
                  R                  :X  a  UR
                  R                  U;   d  [        SU S35      eUS-  n[        R                  " U5      (       aX  [        R                  " U5      n[        R                  " U5      n[        R                  " U5      n[        R                  " U5      nU(       a.  UR                  UR                  5       5      R                  5       nOUR                  U5      R                  5       nU
(       a£  UR                  [        R                  " UR                  S5      UU5      5        UR                  [        R                  " UR!                  S5      UU5      5        UR                  [        R                  " UR#                  S5      SU5      5        O<UUUR                  S5      '   UUUR!                  S5      '   SUUR#                  S5      '   UR%                  U5      R'                  XV5        UR                  [        R(                  S9nU
(       a7  UR                  [        R                  " UR#                  U5      SU5      5        OSUUR#                  U5      '   UR+                  UR                  5       USS9  UR                  U5        GM!     g )NúIIf capturable=True, params and state_steps must be on supported devices: Ú.r   r   rE   éÿÿÿÿ©Úvalue)Ú	enumerater:   ÚcompilerÚis_compilingr   r2   ÚtypeÚAssertionErrorrJ   Úview_as_realÚmulÚcloneÚsignÚcopy_ÚwhereÚgtÚltÚeqÚmul_Úclamp_rP   Úaddcmul_)r   rS   rT   r   rU   rZ   r[   r\   r]   r   r   r   rV   ÚiÚparamrI   rG   rH   r/   Úcapturable_supported_devicesr{   s                        r+   Ú_single_tensor_rpropr‡   ß   se  € ô  ˜f×%‰ˆØ‰xˆÞ#‰t¨$¨ˆØ‰xˆØ‘Mˆ	Ø‰~ˆô �~‰~×*Ñ*×,Ñ,¶Ü+LÓ+NÐ(à—‘×!Ñ! T§[¡[×%5Ñ%5Ó5Ø—L‘L×%Ñ%Ð)EÓEä$Ø_Ð`|Ð_}Ð}~Ðóð ð 	�‰	ˆä×Ò˜E×"Ñ"Ü×%Ò% dÓ+ˆDÜ×%Ò% dÓ+ˆDÜ×&Ò& uÓ-ˆEÜ×*Ò*¨9Ó5ˆIÞØ—8‘8˜DŸJ™J›LÓ)×.Ñ.Ó0‰Dà—8‘8˜D“>×&Ñ&Ó(ˆDæØ�J‰J”u—{’{ 4§7¡7¨1£:¨w¸Ó=Ô>Ø�J‰J”u—{’{ 4§7¡7¨1£:¨x¸Ó>Ô?Ø�J‰J”u—{’{ 4§7¡7¨1£:¨q°$Ó7Õ8à&ˆD�—‘˜“ÑØ'ˆD�—‘˜“ÑØ ˆD�—‘˜“Ñð 	�‰�tÓ×#Ñ# MÔAð �z‰z¬×(=Ñ(=ˆzÐ>ˆÞØ�J‰J”u—{’{ 4§7¡7¨8Ó#4°a¸Ó>Õ?à&'ˆD�—‘˜Ó"Ñ#ð 	�‰�t—y‘y“{ I°RˆÑ8Ø�
‰
�4×òi &r-   c          
      ó   ^• [        U 5      S:X  a  g U(       a  [        S5      e[        R                  R	                  5       (       dB  U
(       a;  [        5       m[        U4S j[        XSS9 5       5      (       d  [        ST S35      e[        R                  " XX#U/5      nUR                  5        GHS  u  u  nnnnnn[        [        [           U5      n[        [        [           U5      n[        [        [           U5      n[        [        [           U5      n[        [        [           U5      n[        R                  R	                  5       (       d>  US   R                  (       a*  [        R                  " U[        R                   " SS	S
9SS9  O[        R                  " US5        U(       a  [#        UUUU5        [        R$                  " UU5      nU	(       a  [        R&                  " U5        [        R(                  " UU5        U	(       a  [        R&                  " U5        Un[        R*                  " U5        U
(       a¬  U H¥  nUR-                  [        R.                  " UR1                  S5      UU5      5        UR-                  [        R.                  " UR3                  S5      UU5      5        UR-                  [        R.                  " UR5                  S5      SU5      5        M§     OEU H?  nUUUR1                  S5      '   UUUR3                  S5      '   SUUR5                  S5      '   MA     [        R6                  " UU5        U H  nUR9                  XV5        M     [        U5      n[;        [        U5      5       HB  nUU   R-                  [        R.                  " UU   R5                  U5      SUU   5      5        MD     AU Vs/ s H  nUR=                  5       PM     nn[        R>                  " UUUSS9  GMV     g s  snf )Nr   z#_foreach ops don't support autogradc              3   óÂ   >#   • U  HT  u  pUR                   R                  UR                   R                  :H  =(       a    UR                   R                  T;   v •  MV     g 7frd   )r2   rv   )Ú.0r?   r/   r†   s      €r+   Ú	<genexpr>Ú&_multi_tensor_rprop.<locals>.<genexpr>?  sO   øé € ð 
ò A‘�ð �H‰H�M‰M˜TŸ[™[×-Ñ-Ñ-÷ >Ø—‘—‘Ð!=Ñ=ô>â@ùs   ƒAAT)Ústrictrn   ro   r"   Úcpu)r2   )Úalphar   rp   rq   ) r9   rw   r:   rt   ru   r   ÚallÚzipr   Ú"_group_tensors_by_device_and_dtypeÚvaluesr   Úlistr   Úis_cpuÚ_foreach_add_r=   r   Ú_foreach_mulÚ_foreach_neg_Ú_foreach_copy_Ú_foreach_sign_r|   r}   r~   r   r€   Ú_foreach_mul_r‚   Úranger{   Ú_foreach_addcmul_) r   rS   rT   r   rU   rZ   r[   r\   r]   r   r   r   rV   Úgrouped_tensorsÚgrouped_params_Úgrouped_grads_Úgrouped_prevs_Úgrouped_step_sizes_Úgrouped_state_steps_Ú_Úgrouped_paramsÚgrouped_gradsÚgrouped_prevsÚgrouped_step_sizesÚgrouped_state_stepsÚsignsr{   rH   r„   rI   Ú
grad_signsr†   s                                   @r+   Ú_multi_tensor_rpropr¬   &  s]  ø€ ô  ˆ6ƒ{�aÓØæÜÐBÓCÐCô �>‰>×&Ñ&×(Ñ(®ZÜ'HÓ'JÐ$Üô 
ô ˜v¸4Ò@ó
÷ 
ñ 
ô
 !Ø[Ð\xÐ[yÐyzÐ{óð ô  ×BÒBØ	˜¨;Ð7ó€Oð ×"Ñ"×$ñ		ñ 	ØØØØØØÜœd¤6™l¨OÓ<ˆÜœT¤&™\¨>Ó:ˆÜœT¤&™\¨>Ó:ˆÜ!¤$¤v¡,Ð0CÓDÐÜ"¤4¬¡<Ð1EÓFÐô �~‰~×*Ñ*×,Ñ,Ð1DÀQÑ1G×1N×1NÜ×ÒØ#¤U§\¢\°#¸eÑ%DÈCóô ×ÒÐ 3°QÔ7ö ÜØ ¨}Ð>Pôô ×"Ò" =°-Ó@ˆÞÜ×Ò Ô&ô
 	×Ò˜]¨MÔ:ÞÜ×Ò Ô.Ø%ˆä×Ò˜UÔ#ÞÛ�Ø—
‘
œ5Ÿ;š; t§w¡w¨q£z°7¸DÓAÔBØ—
‘
œ5Ÿ;š; t§w¡w¨q£z°8¸TÓBÔCØ—
‘
œ5Ÿ;š; t§w¡w¨q£z°1°dÓ;Ö<ò ó
 �Ø#*��T—W‘W˜Q“ZÑ Ø#+��T—W‘W˜Q“ZÑ Ø#$��T—W‘W˜Q“ZÓ ñ ô 	×ÒÐ.°Ô6Û+ˆIØ×Ñ˜]Ö:ñ ,ô
 ˜]Ó+ˆÜ”s˜=Ó)Ö*ˆAØ˜!Ñ×"Ñ"Ü—’˜E !™HŸK™K¨Ó1°1°mÀAÑ6FÓGöñ +ð ñ /<Ó<ªm d�d—i‘i–k©mˆ
Ð<Ü×ÒØ˜JÐ(:À"õ	
òE %ùòB =s   ÏP)Úsingle_tensor_fnr   c
                ó¨  • [         R                  R                  5       (       d"  [        S U 5       5      (       d  [	        S5      eUc  [        XSS9u  påU(       a.  [         R                  R                  5       (       a  [	        S5      eU(       a*  [         R                  R                  5       (       d  [        nO[        nU" U UUUUU
UUUUUUU	S9  g)zhFunctional API that performs rprop algorithm computation.

See :class:`~torch.optim.Rprop` for details.
c              3   óV   #   • U  H  n[        U[        R                  5      v •  M!     g 7frd   )r#   r:   r   )rŠ   Úts     r+   r‹   Úrprop.<locals>.<genexpr>¶  s!   é € ð 5Ú-8¨Œ
�1”e—l‘l×#Ð#ª[ùs   ‚')zPAPI has changed, `state_steps` argument must contain a list of singleton tensorsNF)Ú	use_fusedz6torch.jit.script not supported with foreach optimizers)rZ   r[   r\   r]   r   r   r   rV   )
r:   rt   ru   r�   rM   r   ÚjitÚis_scriptingr¬   r‡   )r   rS   rT   r   rU   r   r   r   r   rV   rZ   r[   r\   r]   r¤   Úfuncs                   r+   r   r   œ  sË   € ô4 �>‰>×&Ñ&×(Ñ(´ñ 5Ù-8ó5÷ 2ñ 2ô Ø^ó
ð 	
ð �Ü1Ø¨eñ
‰
ˆö ”5—9‘9×)Ñ)×+Ñ+ÜÐSÓTÐTæ”u—y‘y×-Ñ-×/Ñ/Ü"‰ä#ˆáØØØØØØ#Ø#ØØØØØ%Øór-   )NFFFF)Ú__doc__Útypingr   r:   r   Ú	optimizerr   r   r   r	   r
   r   r   r   r   r   r   r   r   r   Ú__all__r   r”   r<   rj   r‡   r¬   r   rD   r-   r+   Ú<module>rº      sÏ  ðá 8å ã Ý ÷÷ ÷ ÷ ð$ �GÐ
€ôHˆIô HðX!LðD	ð 
ˆð 	ð 
Ðð 	Ø	ˆð 	Ø	ˆð 	Ø	Ðð ðñE1ð „ðlDØ�‰LðDà�‰<ðDð �‰<ðDð �V‘ð	Dð
 �f‘ðDð ðDð ðDð ðDð ðDð ðDð ðDð ðDð ðDð 
ôDðNo
Ø�‰Lðo
à�‰<ðo
ð �‰<ðo
ð �V‘ð	o
ð
 �f‘ðo
ð ðo
ð ðo
ð ðo
ð ðo
ð ðo
ð ðo
ð ðo
ð ðo
ð 
ôo
ñl  Ð1EÑFð  ØØØ Øñ;Ø�‰Lð;à�‰<ð;ð �‰<ð;ð �V‘ð	;ð
 �f‘ð;ð �D‰[ð;ð ð;ð ð;ð ð;ð ð;ð ð;ð ð;ð  ð!;ð" ð#;ð$ 
ô%;ó Gñ;r-   