"Eiah(PSrSSKJr SSKrSSKJr SSKJrJrJrJ r J r J r J r J r JrJrJrJrJrJrJrJr SS/r"S S\5rS S \S \ S \S \S \S3 -\lS\\S\\S\\S\\S\\S\\S\S\S\S\S\S\S\S\S\S\S\S S4$S!jrS\\S\\S\\S\\S\\S\\S\S\S\S\S\S\S\S\S\S\S\S S4$S"jr\ "\S#9S&S\\S\\S\\S\\S\\S\\S\S$\S-S\S\S\S\S\S\S\S\S\S\S S4&S%jj5rg)'z'Implementation for the NAdam algorithm.)castN)Tensor)_capturable_doc_default_to_fused_or_foreach_differentiable_doc_disable_dynamo_if_unsupported _foreach_doc!_get_capturable_supported_devices_get_scalar_dtype _get_value _maximize_doc _params_doc_stack_if_compiling _to_scalar_use_grad_for_differentiable _view_as_real OptimizerParamsTNAdamnadamc^\rSrSrSSSSSS.S\S\\-S\\\4S\S \S \S \S \S-S \S\S\SS4U4Sjjjjr U4Sjr Sr \ SSj5r SrU=r$)r FN)foreachmaximize capturabledifferentiableparamslrbetaseps weight_decaymomentum_decaydecoupled_weight_decayrrrrreturnc >[U[5(aUR5S:wa [S5eSU::d[SU35eSU::d[SU35eSUSs=::aS:dO [SUS35eSUSs=::aS:dO [S US35eSU::d[S U35eSU::d[S U35eUUUUUUU UU U S . n [T U]X5 g) NrzTensor lr must be 1-elementzInvalid learning rate: zInvalid epsilon value: r?z#Invalid beta parameter at index 0: z#Invalid beta parameter at index 1: zInvalid weight_decay value: zInvalid momentum_decay value: ) rr r!r"r#r$rrrr) isinstancernumel ValueErrorsuper__init__)selfrrr r!r"r#r$rrrrdefaults __class__s N/home/mande/repo/quber/.venv/lib/python3.13/site-packages/torch/optim/nadam.pyr-NAdam.__init__!s b& ! !bhhjAo:; ;by6rd;< <cz6se<= =eAh$$B58*MN NeAh$$B58*MN Nl";L>JK Kn$=n=MNO O(,&< $,   *cb>[TU]U5 URGHnURSS5 URSS5 URSS5 URSS5 URSS5 USGH&nURR U/5n[ U5S:wdM1[R"US 5(d`[US 5nUS(a([R"U[5URS 9O[R"U[5S 9US '[R"US 5(aMUS nUS(a([R"U[5URS 9O[R"U[5S 9US 'GM) GM g) NrFrrrr$rrstepdtypedevicer7 mu_product) r, __setstate__ param_groups setdefaultstategetlentorch is_tensorfloattensorr r8)r.r>grouppp_statestep_val mu_prod_valr0s r1r;NAdam.__setstate__Lsj U#&&E   Z /   Y -   \5 1   -u 5   5u =8_**..B/w<1$ ??76?;;#(#9 %\2"LL (0A0CAHH"'h>O>Q!R  !??7<+@AA&-l&; %\2"LL +3D3Fqxx"'kARAT!U  -% 'r3cSnUSGHn U RcMU[R"U 5-nURU 5 U RR(a [ S5eURU R5 UR U n [U 5S:XaUS(a([R"S[5U RS9O[R"S[5S 9U S 'US(a([R"S[5U RS9O[R"S [5S 9U S '[R"U [RS 9U S'[R"U [RS 9U S'URU S5 URU S5 URU S 5 URU S 5 GM U$)NFrz'NAdam does not support sparse gradientsrrr6r'r9r5r(r:) memory_formatexp_avg exp_avg_sq)gradrA is_complexappend is_sparse RuntimeErrorr>r@zerosr r8rDones zeros_likepreserve_format) r.rEparams_with_gradgradsexp_avgs exp_avg_sqs mu_products state_steps has_complexrFr>s r1 _init_groupNAdam._init_groupjs xAvv!u//22  ''*66##&'PQQ QVV$ 1 u:? !. B.?.A!((S"\\#5F5HI&M!. 2->-@R"\\#5F5HI,' (-'7'7)>)>(E)$+0*:*:)>)>+E,'i 01""5#67""5#67""5=1I!Jr3cUR5 SnUb%[R"5 U"5nSSS5 URH}n/n/n/n/n/n/n [ [ [ [ 4US5upURUUUUUUU 5n [UUUUUU U U USUSUSUSUSUSUS US US U S 9 M U$!,(df  N=f) zPerform a single optimization step. Args: closure (Callable, optional): A closure that reevaluates the model and returns the loss. Nr rr"r#r!rr$rrr) beta1beta2rr"r#r!rr$rrrr_) '_accelerator_graph_capture_health_checkrA enable_gradr<rtuplerCr`r) r.closurelossrErYrZr[r\r]r^rcrdr_s r1r5 NAdam.steps$ 446  ""$y%&&E-/ "$E%'H(*K(*K(*KeUl 3U7^DLE** K  ;">2$%56%Lz*',-E'Fi( .$%56'% ''P W%$s C  CrL)gMb`?)g?g+?g:0yE>rgMbp?FN)__name__ __module__ __qualname____firstlineno__rrCrrgboolr-r;r`rr5__static_attributes__ __classcell__)r0s@r1rr s"%1 $',)+ $ $)+)+ FN)+UE\" )+  )+  )+)+!%)+)+)+)+)+ )+)+V<0d"6"6r3a Implements NAdam algorithm. .. math:: \begin{aligned} &\rule{110mm}{0.4pt} \\ &\textbf{input} : \gamma_t \text{ (lr)}, \: \beta_1,\beta_2 \text{ (betas)}, \: \theta_0 \text{ (params)}, \: f(\theta) \text{ (objective)} \\ &\hspace{13mm} \: \lambda \text{ (weight decay)}, \:\psi \text{ (momentum decay)} \\ &\hspace{13mm} \: \textit{decoupled\_weight\_decay}, \:\textit{maximize} \\ &\textbf{initialize} : m_0 \leftarrow 0 \text{ ( first moment)}, v_0 \leftarrow 0 \text{ ( second moment)} \\[-1.ex] &\rule{110mm}{0.4pt} \\ &\textbf{for} \: t=1 \: \textbf{to} \: \ldots \: \textbf{do} \\ &\hspace{5mm}\textbf{if} \: \textit{maximize}: \\ &\hspace{10mm}g_t \leftarrow -\nabla_{\theta} f_t (\theta_{t-1}) \\ &\hspace{5mm}\textbf{else} \\ &\hspace{10mm}g_t \leftarrow \nabla_{\theta} f_t (\theta_{t-1}) \\ &\hspace{5mm} \theta_t \leftarrow \theta_{t-1} \\ &\hspace{5mm} \textbf{if} \: \lambda \neq 0 \\ &\hspace{10mm}\textbf{if} \: \textit{decoupled\_weight\_decay} \\ &\hspace{15mm} \theta_t \leftarrow \theta_{t-1} - \gamma \lambda \theta_{t-1} \\ &\hspace{10mm}\textbf{else} \\ &\hspace{15mm} g_t \leftarrow g_t + \lambda \theta_{t-1} \\ &\hspace{5mm} \mu_t \leftarrow \beta_1 \big(1 - \frac{1}{2} 0.96^{t \psi} \big) \\ &\hspace{5mm} \mu_{t+1} \leftarrow \beta_1 \big(1 - \frac{1}{2} 0.96^{(t+1)\psi}\big)\\ &\hspace{5mm}m_t \leftarrow \beta_1 m_{t-1} + (1 - \beta_1) g_t \\ &\hspace{5mm}v_t \leftarrow \beta_2 v_{t-1} + (1-\beta_2) g^2_t \\ &\hspace{5mm}\widehat{m_t} \leftarrow \mu_{t+1} m_t/(1-\prod_{i=1}^{t+1}\mu_i)\\[-1.ex] & \hspace{11mm} + (1-\mu_t) g_t /(1-\prod_{i=1}^{t} \mu_{i}) \\ &\hspace{5mm}\widehat{v_t} \leftarrow v_t/\big(1-\beta_2^t \big) \\ &\hspace{5mm}\theta_t \leftarrow \theta_t - \gamma \widehat{m_t}/ \big(\sqrt{\widehat{v_t}} + \epsilon \big) \\ &\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 `Incorporating Nesterov Momentum into Adam`_. z Args: a lr (float, Tensor, optional): learning rate (default: 2e-3) betas (Tuple[float, float], optional): coefficients used for computing running averages of gradient and its square (default: (0.9, 0.999)) eps (float, optional): term added to the denominator to improve numerical stability (default: 1e-8) weight_decay (float, optional): weight decay (L2 penalty) (default: 0) momentum_decay (float, optional): momentum momentum_decay (default: 4e-3) decoupled_weight_decay (bool, optional): whether to decouple the weight decay as in AdamW to obtain NAdamW. If True, the algorithm does not accumulate weight decay in the momentum nor variance. (default: False) z z .. _Incorporating Nesterov Momentum into Adam: https://openreview.net/forum?id=OM0jvwB8jIp57ZJjtNEZ .. _Decoupled Weight Decay Regularization: https://arxiv.org/abs/1711.05101 rrZr[r\r]r^rcrdrr"r#r!r$rrrr_r%c v[RR5(d [U5n[ U5GHunnU (dUUOUU*nUUnUUnUUnUUn[R "U5(aX[R "U5n[R "U5n[R "U5n[R "U5n[RR5(dU(a[5nURRURRs=:XaURR:XaO OURRU;d[SUS35eUS- nU(aUnO [U5nSUU-- nU S:wa.U (aURSX-- 5 OURUU S9nUSSSUU ---- -nUSSSUS-U ---- -nUU-nUR!USU- 5 URU5R#UUSU- S 9 UR%U5R'5nU(dU(a^URU 5nUU-nUU*SU- -SU- - -nUU*U-SU- - -nUR)UU5 UR)UU5 GM[U5U-nUR+U 5 UR)UUU*SU- -S[U5- - S 9 UR)UU[-[.U*U-SU- - 5S 9 GM g) NzVIf capturable=True, params, mu_products and state_steps must be on supported devices: .rralphar(?Q?)value)rAjit is_scriptingr enumeraterQ view_as_realcompiler is_compilingr r8typeAssertionErrorr mul_addlerp_addcmul_divsqrtaddcdiv_add_rrC)rrZr[r\r]r^rcrdrr"r#r!r$rrrr_iparamrPrNrOr:step_tcapturable_supported_devicesr5bias_correction2mumu_nextdenommu_product_nexts r1_single_tensor_nadamrs&( 99 ! ! # # ^f%5'uQxeAhY1+ ^  ^ Q   E " "&&u-E%%d+D((1G++J7J~~**,,+L+N ( !!Z%6%6%;%;Qv}}?Q?QQLL%%)EE$--I,J!M !  Df%Dud{? 1 % 1r001xx\x:cC4D>,A#BCC D3$(n1L(M!NNO b   dAI&''d!e)'D/0557 ZIIcNE)72OB3#(+sZ/?@AD"w#2G!HIG NN4 ' NN7E *(4w>O JJsO NNeRC38$4j>T8T$U   NN5B3=S?5J"KL  M&r3c  ^,[U5S:XagU(a [S5e[RR 5(dBU(a;[ SS9m,[ U,4Sj[XUSS955(d[ST,S 35e[U5n[R"XX#XE/5nUR5GHuunnnnnnn[[[U5n[[[U5n[[[U5n[[[U5n[[[U5n[[[U5nU(a[UUUU5 U (a[R "U5n[RR 5(d>USR"(a*[R$"U[R&"S S S 9S S 9 O[R$"US5 U S:waXU (a[R("USX-- 5 O4U (a[R$"UUU S 9 O[R*"UUU S 9n[R,"UUSU- 5 [R("UU5 [R."UUUSU- 5 [R0"U5nU(GaB[R2"UU 5n [R4"SU 5n![R("U!S5 [R$"U!S 5 [R("U!U5 [R$"U U 5 [R4"SU 5n"[R("U"S5 [R$"U"S 5 [R("U"U5 A [R4"UU5n#[R6"U#S 5 [R8"U#5 [R:"U#5 O{UV$s/sHn$SU[=U$5-- S-PM n#n$UV$s/sHn$US SS[=U$5U ---- -PM n!n$UV$s/sH n$US SS[=U$5S-U ---- -PM" n"n$[R("UU!5 [R>"UU#5 [R$"UU 5 A#U(Ga[R6"U!S 5 [R("U!U5 [R@"US 5n%[R8"U%5 [R>"U!U%5 U!n&A%[R2"UU"5n%[R("U"U5 [R6"U%S 5 [R>"U"U%5 U"n'A%[R2"U&U5n([R."U(U'U5 [RB"UU(U5 GM5[E[UU!SS9V)V*s/sH)un)n*[=U5S U*- -S [=U)5- - S-PM+ sn*n)5n&[E[UU"SS9V)V+s/sH)un)n+[=U5U+-S [=U)5U+-- - S-PM+ sn+n)5n'[RB"UUUU&5 [RB"UUUU'5 GM gs sn$fs sn$fs sn$fs sn*n)fs sn+n)f)Nrz#_foreach ops don't support autogradF) supports_xlac3 ># UHyupnURRURRs=:H=(a URR:HOs =(a URRT;v M{ g7frk)r8r).0rFmpr5rs r1 &_multi_tensor_nadam..sa  R t HHMMRYY^^ ? ?t{{/?/? ? > !== >QsBBT)strictzWIf capturable=True, params, mu_products, and state_steps must be on supported devices: rtr(cpu)r8rurrxgrw)#r@rrAr~rr allziprr"_group_tensors_by_device_and_dtypevaluesrlistrr _foreach_negis_cpu _foreach_add_rD _foreach_mul_ _foreach_add_foreach_lerp__foreach_addcmul_ _foreach_sqrt _foreach_mul _foreach_pow _foreach_sub_ _foreach_neg__foreach_sqrt_r _foreach_div_ _foreach_sub_foreach_addcdiv_r)-rrZr[r\r]r^rcrdrr"r#r!r$rrrr_grouped_tensorsgrouped_params_grouped_grads_grouped_exp_avgs_grouped_exp_avg_sqs_grouped_mu_products_grouped_state_steps__grouped_params grouped_gradsgrouped_exp_avgsgrouped_exp_avg_sqsgrouped_mu_productsgrouped_state_stepsexp_avg_sq_sqrtexponentmusmu_nextsbias_correction_sqrtr5rstep_size_gradsstep_size_expavg numeratorr:rrrs- @r1_multi_tensor_nadamr}sp( 6{aBCC >> & & ( (Z'H( $  #6 DQ   !V/03  BBBB {HO  " " $  d6lO<T&\>: V .?@"4<1EF"4<1EF"4<1EF   /?AT  !..}=M ~~**,,1DQ1G1N1N   #U\\#e%DC     3Q 7 1 %##NA8I4IJ''%~\%*$6$6%~\%M -}a%iH /7   q5y  --.AB ))*=~NH$$T84C   T *   S )   U +   . 9))$9H   $ /   # .   % 0#(#5#5e=P#Q    4c :    4 5  !5 6DW$CV4Uj...36CV !$ 0/Dsdz$/?./P&QRRS/ 00D*T*:Q*>.)P QRRT/  /5 O-AB OS1 !    S )   R (&&':C@E    &   U +!O&&':HEE   " -   s +   % 0' **?MJI  # #I/?AQ R  # #NI O1+..A3t*T*T B ^sRx0C*Z:P4PQUWW*TO 303+Xd0 0+ G #2!"J!7'!AAC  0     # #    # #   C %`$b s[3$['[ 0[ 0[ )single_tensor_fnrc [SU55(d [S5e[SU55(d [S5eUc [X SS9unnU(a.[RR 5(a [S5eU(a*[RR 5(d[ nO[nU"UUUUUUU U UUUU UUUU U S 9 g) zhFunctional API that performs NAdam algorithm computation. See :class:`~torch.optim.NAdam` for details. c3V# UHn[U[R5v M! g7frkr)rArrts r1rnadam..x@Kqz!U\\**K')zPAPI has changed, `state_steps` argument must contain a list of singleton tensorsc3V# UHn[U[R5v M! g7frkrrs r1rr}rrzPAPI has changed, `mu_products` argument must contain a list of singleton tensorsNF) use_fusedz6torch.jit.script not supported with foreach optimizers) rcrdrr"r#rr$r!rrr_)rrTrrArzr{rr)rrZr[r\r]r^r$rrrr_rrcrdrr"r#r!rfuncs r1rr\s8 @K@ @ @ ^   @K@ @ @ ^  1 e 7599))++STTuyy--//"#  !%5 %#r3)FNFFFF)__doc__typingrrAr optimizerrrrr r r r r rrrrrrrr__all__rrrCrprrrrLr3r1rsu. ( G sIsn&N       !O> Fa La <a6laf a f a f a a a aaa a!aa !a"#a$%a& 'aH\ L\ <\6l\f \ f \ f \ \ \ \\\ \!\\ !\"#\$%\& '\~ 1EF$) D LD <D6lDf D f D f D!DD[DDDDD !D" #D$ %D&'D()D* +D, -DGDr3