+ &jA ^RIHtHt^RIt^RIHt^RIHtHtHtH t H t H t H t H t HtHtHtHtHtHtRR.t!RR]4tRR ] R ] R ] R ] R ] R 2 ,]nR RltRRlt] !]R7RRRll4tR#))AnycastN)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 OptimizerParamsTAdadeltaadadeltacaa]tRt^toR RRRRRR/V3RlV3RlllltV3RltV3RlR lt]R R l4tR t Vt V;t #)r capturableFmaximizedifferentiablecr<V^8dQhRS[RS[S[,RS[RS[RS[RS[R,RS[R S[R S[R R/ #) paramslrrhoeps weight_decayforeachNrrrreturn)rfloatrbool)format __classdict__s"l/Users/jameslopez/projects/CWCArchive/cwc-podcast/.venv/lib/python3.14/site-packages/torch/optim/adadelta.py __annotate__Adadelta.__annotate__sw"+"+"+ FN"+ "+  "+  "+"+"+"+"+ "+cr<\V\4'd!VP4^8wd \R4hRV8:g\RV 24hRTu;8:dR8:gM\RV 24hRV8:g\RV 24hRV8:g\RV 24hRVR VR VR VR VR VRVRV /p \S V`W4R#)zTensor lr must be 1-elementgzInvalid learning rate: ?zInvalid rho value: zInvalid epsilon value: zInvalid weight_decay value: rrrr rrr!rN) isinstancernumel ValueErrorsuper__init__) selfrrrrr r!rrrdefaults __class__s &&&&&&&$$$ r'r2Adadelta.__init__s b& ! !bhhjAo:; ;by6rd;< <c S 23%89 9cz6se<= =l";L>JK K " 3 3 L  * w n   *r*c<\SV`V4VPEFpVPRR4VPRR4VPRR4VPRR4VR,FpVPP V.4p\ V4^8wgK1\P!VR,4'dKV\VR,4pVR,'d,\P!V\4VPR7M\P!V\4R 7VR&K EK! R#) r!NrFrrrstepdtypedevicer:) r1 __setstate__ param_groups setdefaultstategetlentorch is_tensorr#tensorr r;)r3r@grouppp_statestep_valr5s&& r'r=Adadelta.__setstate__As U#&&E   Y -   Z /   -u 5   \5 18__**..B/w<1$U__WV_-M-M$WV_5H !.. $,=,?#\\(:K:MN FO % 'r*c <V^8dQhRS[S[S[3,RS[S[,RS[S[,RS[S[,RS[S[,RS[S[,/#)rrFparams_with_gradgrads square_avgs acc_deltas state_steps)dictstrrlistr)r%r&s"r'r(r)Tsc))CH~)v,)F| ) &\ ) L )&\)r*cbRpVR,EFpVPfKV\P!V4,pVPV4VPP'd \ R4hVPVP4VP V,p \V 4^8XdVR,'d,\P!R \4VPR7M\P!R \4R7V R&\P!V\PR7V R&\P!V\PR7V R &VPV R,4VPV R ,4VPV R,4EK V#) Frz*Adadelta does not support sparse gradientsrr9r<r8) memory_format square_avg acc_delta) gradrC is_complexappend is_sparse RuntimeErrorr@rBzerosr r; zeros_likepreserve_format) r3rFrLrMrNrOrP has_complexrGr@s &&&&&&& r' _init_groupAdadelta._init_groupTsV xAvv~ 5++A. .K  # #A &vv"#OPP LL JJqME5zQ\**KK*;*=ahhOR/@/BCf ',&6&6U%:%:'l#&+%5%5U%:%:&k"   u\2 3   eK0 1   uV} -9!<r*c VP4RpVe.\P!4;_uu_4V!4pRRR4VPFp.p.p.p.p.pVR,VR,VR,VR,VR,VR,VR,VR ,3wp p p p p pppVP W4WVWx4p\ VVVVVV V V V V VVVVR 7K V# +'giL;i) zPerform a single optimization step. Args: closure (Callable, optional): A closure that reevaluates the model and returns the loss. Nrrrr r!rrr) rrrr r!rrrra)'_accelerator_graph_capture_health_checkrC enable_gradr>rbr)r3closurelossrFrLrMrNrOrPrrrr r!rrrras&& r'r8 Adadelta.steps# 446  ""$$y%&&E-/ "$E(*K')J(*Kd e e n%i j!&'l#  **ZK  )!-%' ='^ e%$s C C* rX)r-g?gư>rNN) __name__ __module__ __qualname____firstlineno__r2r=rbrr8__static_attributes____classdictcell__ __classcell__)r5r&s@@r'rrsW"+!"+"+ %"+"+H&))V"="==r*aImplements Adadelta algorithm. .. math:: \begin{aligned} &\rule{110mm}{0.4pt} \\ &\textbf{input} : \gamma \text{ (lr)}, \: \theta_0 \text{ (params)}, \: f(\theta) \text{ (objective)}, \: \rho \text{ (decay)}, \: \lambda \text{ (weight decay)} \\ &\textbf{initialize} : v_0 \leftarrow 0 \: \text{ (square avg)}, \: u_0 \leftarrow 0 \: \text{ (accumulate variables)} \\[-1.ex] &\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}if \: \lambda \neq 0 \\ &\hspace{10mm} g_t \leftarrow g_t + \lambda \theta_{t-1} \\ &\hspace{5mm} v_t \leftarrow v_{t-1} \rho + g^2_t (1 - \rho) \\ &\hspace{5mm}\Delta x_t \leftarrow \frac{\sqrt{u_{t-1} + \epsilon }}{ \sqrt{v_t + \epsilon} }g_t \hspace{21mm} \\ &\hspace{5mm} u_t \leftarrow u_{t-1} \rho + \Delta x^2_t (1 - \rho) \\ &\hspace{5mm}\theta_t \leftarrow \theta_{t-1} - \gamma \Delta x_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 `ADADELTA: An Adaptive Learning Rate Method`_. z Args: ar lr (float, Tensor, optional): coefficient that scale delta before it is applied to the parameters (default: 1.0) rho (float, optional): coefficient used for computing a running average of squared gradients (default: 0.9). A higher value of `rho` will result in a slower average, which can be helpful for preventing oscillations in the learning process. eps (float, optional): term added to the denominator to improve numerical stability (default: 1e-6). weight_decay (float, optional): weight decay (L2 penalty) (default: 0) z zd .. _ADADELTA\: An Adaptive Learning Rate Method: https://arxiv.org/abs/1212.5701 c&V^8dQhR\\,R\\,R\\,R\\,R\\,R\R\R\R \R \R \R \R \RR/#rrrMrNrOrPrrrr rrrrar"NrSrr#r$)r%s"r'r(r(s9%9% L9% <9%f9%V 9% f 9% 9% 9% 9%9%9%9%9%9% 9%r*c a\PP4'gzV 'dr\RR7o\;QJd*V3Rl\ WRR74F 'dK RM RM!V3Rl\ WRR744'g\ RS R24h\PP4'g \V4p\ WW#VRR7EFwrpppV^, pV 'gTMV)pV^8wdVPWR7p\P!V 4'dC\P!V4p\P!V4p\P!V4pVPV4PW^V, R7VPV4P4pVPV4P4pV 'dVP!4pVP#V4PV4VPV4PVV^V, R7\P!V 4'd\P$!V4pV P'VV)R7EK R #) F supports_xlac3<"TFUwrVPPVPP8H;'dVPPS9xKW R#5irjr;type.0rGr8capturable_supported_devicess& r' *_single_tensor_adadelta.. T A HHMMT[[-- - > > !== >@ :A "A TstrictIIf capturable=True, params and state_steps must be on supported devices: .alphavalueN)rCcompiler is_compilingr allzipAssertionErrorjit is_scriptingraddrZ view_as_realmul_addcmul_sqrt_clonediv_view_as_complexadd_)rrMrNrOrPrrrr rrrraparamrYrVrWr8stddeltar}s&&&&&$$$$$$$$ @r'_single_tensor_adadeltars" >> & & ( (Z'H( $s v4@ sss v4@   ![\x[yyz{  99 ! ! # # ^47{ D550ZD  #t$ 1 88E86D   E " "++J7J**95I%%d+D%%dC%@nnS!'') c"((* KKME 3T"s$$UES$A   E " "))%0E 5 $15r*c&V^8dQhR\\,R\\,R\\,R\\,R\\,R\R\R\R \R \R \R \R \RR/#rsrt)r%s"r'r(r(1sgBgB LgB <gBfgBV gB f gB gB gB gBgBgBgBgBgB gBr*c aV 'd \R4h\PP4'gzV 'dr\ RR7o\ ;QJd*V3Rl\ WRR74F 'dK RM RM!V3Rl\ WRR744'g\RS R24h\V4^8XdR#\V4p\P!WW#V.4p V P4EFwwpppppp\\\,V4p\\\,V4p\\\,V4p\\\,V4p\\\,V4pV 'd\VVVV4\PP4'gJV^,P 'd1\P"!V\P$!R R R 7R R 7M\P"!V^4V 'd\P&!V4pV^8wd<V 'd\P"!VVVR 7M\P(!VVVR 7p\P*!VV4\P,!VVV^V, R 7\P(!VV4p\P.!V4\P(!VV4p\P.!V4\P0!VV4\P*!VV4\P*!VV4\P,!VVV^V, R 7V 'dS\3V\P4'd3\P*!VV)4\P"!VV4EK\P"!VVV)R 7EK R#)z#_foreach ops don't support autogradFrvc3<"TFUwrVPPVPP8H;'dVPPS9xKW R#5irjryr{s& r'r~)_multi_tensor_adadelta..IrrTrrrNr-cpu)r;rr)rrCrrr rrrBrr"_group_tensors_by_device_and_dtypevaluesrrSrris_cpu _foreach_add_rE _foreach_neg _foreach_add _foreach_mul__foreach_addcmul__foreach_sqrt_ _foreach_div_r.)rrMrNrOrPrrrr rrrragrouped_tensorsdevice_params_ device_grads_device_square_avgs_device_acc_deltas_device_state_steps__ device_params device_gradsdevice_square_avgsdevice_acc_deltasdevice_state_stepsrdeltasr}s&&&&&$$$$$$$$ @r'_multi_tensor_adadeltar1s BCC >> & & ( (Z'H( $s v4@ sss v4@   ![\x[yyz{  6{a BBBB  =O  " " $   T&\>: DL-8 !$v,0CD f/AB!$v,0CD  |-?AR ~~**,,1CA1F1M1M1M   "ELLU$C3     2A 6  --l;L 1 ##L-|T$11 -|   .4   l!c'   !3S9 S!##$5s; V$ FC( FL1 -s3  166SQ *R66    ,    v 6    vbS A Aq %r*)single_tensor_fnc@V^8dQhR\\,R\\,R\\,R\\,R\\,R\R\R,R \R \R \R \R \R\R\RR/#)rrrMrNrOrPrr!Nrrarrrr rr")rSrr$r#)r%s"r'r(r(s== L= <=f=V = f ==D[=== = = = !="#=$ %=r*c \PP4'gF\;QJdRV4F 'dK RM RM !RV44'g \ R4hVf\ WRR7wrV'd0\P P4'd \ R4hV'd,\P P4'g\pM\pV!VVVVVV V V V V VVVR7 R#) znFunctional API that performs Adadelta algorithm computation. See :class:`~torch.optim.Adadelta` for details. c3V"TFp\V\P4xK! R#5irj)r.rCr)r|ts& r'r~adadelta..s!5-8 1ell##[s')FTzPAPI has changed, `state_steps` argument must contain a list of singleton tensorsN) use_fusedz6torch.jit.script not supported with foreach optimizers)rrrr rrrra) rCrrrr]rrrrr)rrMrNrOrPrr!rrarrrr rrfuncs&&&&&&&&&$$$$$ r'rrs6 >> & & ( (5-855-8522 ^  1 e 599))++STTuyy--//%&   !%r*)FNFF)typingrrrCr optimizerrrrr r r r r rrrrrr__all__r__doc__rrrrXr*r'rs $ z "ayaJ8       90 j9%xgBT 1HI=J=r*