o 闦iLC @sddlmZmZmZmZmZddlZddlmZddlm Z m Z m Z m Z m Z mZmZmZmZmZmZmZmZmZddgZGdddeZd d ed e d ed e d e d e_deedeedeedeedeedededededededededefddZdeedeedeedeedeedededededededededefddZe ed   ! ! ! !d$deedeedeedeedeed"eedededededededededefd#dZdS)%)castListOptionalTupleUnionN)Tensor)_capturable_doc_default_to_fused_or_foreach_differentiable_doc_disable_dynamo_if_unsupported _foreach_doc!_get_capturable_supported_devices_get_scalar_dtype _get_value _maximize_doc _params_doc_use_grad_for_differentiable _view_as_real OptimizerParamsTAdamaxadamaxcseZdZ     dddddded eeefd eeefd ed ed ee de de de ffddZ fddZ ddZ e dddZZS)rMb`?g?g+?:0yE>rNF)maximizedifferentiable capturableparamslrbetaseps weight_decayforeachrrrc st|tr|dkrtdd|kstd|d|ks%td|d|dkr1dks;ntd|dd|dkrGdksQntd |dd|ks\td |t|||||||| d } t|| dS) NrzTensor lr must be 1-elementzInvalid learning rate: zInvalid epsilon value: r?z#Invalid beta parameter at index 0: z#Invalid beta parameter at index 1: zInvalid weight_decay value: )r r!r"r#r$rrr) isinstancernumel ValueErrordictsuper__init__) selfrr r!r"r#r$rrrdefaults __class__P/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/torch/optim/adamax.pyr,s.  zAdamax.__init__cst||jD]S}|dd|dd|dd|dd|dD]4}|j|g}t|dkr[t|ds[t |d}|drQtj |t |j d ntj |t d |d<q'q dS) Nr$rFrrrrstepdtypedevicer5) r+ __setstate__ param_groups setdefaultstategetlentorch is_tensorfloattensorrr6)r-r;grouppp_statestep_valr/r1r2r8Ds&         zAdamax.__setstate__c Csd}|dD]n}|jdurq|t|O}|||jjr"td||j|j|} t| dkr_|drAtjdt |j dntj dt d | d <tj |tj d | d <tj |tj d | d <|| d || d || d q|S)NFrz(Adamax does not support sparse gradientsrrr1r4r%r7r3) memory_formatexp_avgexp_inf)gradr> is_complexappend is_sparse RuntimeErrorr;r=zerosrr6rA zeros_likepreserve_format) r-rBparams_with_gradgradsexp_avgsexp_infs state_steps has_complexrCr;r1r1r2 _init_groupWs2        zAdamax._init_groupcCs|d}|dur!t |}Wdn1swY|jD]K}g}g}g}g}g}|d\} } |d} |d} |d} |d}|d}|d}|d }|||||||}t|||||| | | | | |||||d q$|S) zPerforms a single optimization step. Args: closure (Callable, optional): A closure that reevaluates the model and returns the loss. Nr!r"r r#r$rrr) r"beta1beta2r r#r$rrrrV) _cuda_graph_capture_health_checkr> enable_gradr9rWr)r-closurelossrBrQrRrSrTrUrXrYr"r r#r$rrrrVr1r1r2r3zsR    z Adamax.step)rrrrNN)__name__ __module__ __qualname__rrr@rrrboolr,r8rWrr3 __classcell__r1r1r/r2rsB       &#aImplements Adamax algorithm (a variant of Adam based on infinity norm). .. math:: \begin{aligned} &\rule{110mm}{0.4pt} \\ &\textbf{input} : \gamma \text{ (lr)}, \beta_1, \beta_2 \text{ (betas)},\theta_0 \text{ (params)},f(\theta) \text{ (objective)}, \: \lambda \text{ (weight decay)}, \\ &\hspace{13mm} \epsilon \text{ (epsilon)} \\ &\textbf{initialize} : m_0 \leftarrow 0 \text{ ( first moment)}, u_0 \leftarrow 0 \text{ ( infinity norm)} \\[-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}m_t \leftarrow \beta_1 m_{t-1} + (1 - \beta_1) g_t \\ &\hspace{5mm}u_t \leftarrow \mathrm{max}(\beta_2 u_{t-1}, |g_{t}|+\epsilon) \\ &\hspace{5mm}\theta_t \leftarrow \theta_{t-1} - \frac{\gamma m_t}{(1-\beta^t_1) u_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 `Adam: A Method for Stochastic Optimization`_. 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 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) z zd .. _Adam\: A Method for Stochastic Optimization: https://arxiv.org/abs/1412.6980 rrRrSrTrUr"rXrYr r#rrrrVc Cst|D]\}}||}| s|n| }||}||}||}tjs?| r?t}|jj|jjkr7|jj|vs?Jd|d|d7}| dkrN|j|| d}t|rgt |}t |}t |}t |}| |d|| stj | || ||dn!t| |d| |dgd}|tj|ddd| r||d}||||}|||qd|t|}||}|j||| d qdS) NIIf capturable=True, params and state_steps must be on supported devices: .rralpha)outF)keepdim)value) enumerater>compiler is_compilingrr6typeaddrJ view_as_reallerp_maximummul_absadd_cat unsqueeze unsqueeze_copy_amaxdiv_addcdiv_r)rrRrSrTrUr"rXrYr r#rrrrViparamrIrGrHstep_tcapturable_supported_devicesnorm_bufneg_bias_correctiondenombias_correctionclrr1r1r2_single_tensor_adamaxsR       "  rc s4| rJdt|dkrdStjs0| r0tddtfddt||Ds0Jddt|||||g}| D]\\}}}}}}t t t |}t t t |}t t t |}t t t |}t t t |}| rst ||||| rzt|}tjs|djrtj|tjd d d d d nt|d | dkr| rtj||| d ntj||| d }t||d t||| s| dkrt|}nt|t||t||| rt|}t|d t|t||}t|||q>fdd|D}fdd|D}t||||q>dS)Nz#_foreach ops don't support autogradrF) supports_xlac3s0|]\}}|jj|jjko|jjvVqdSr^)r6rn).0rCr3)rr1r2 Gs   z'_multi_tensor_adamax..rdrer&cpu)r6rfrcsg|] }dt|qS)rr)rr3)rXr1r2 sz(_multi_tensor_adamax..csg|] }t|dqS)r)rbc)r r1r2rs)r=r>rlrmrallzipr"_group_tensors_by_device_and_dtypevaluesrrrr _foreach_negis_cpu _foreach_add_rA _foreach_add_foreach_lerp_ _foreach_mul_ _foreach_abs _foreach_abs__foreach_maximum_ _foreach_pow _foreach_sub_ _foreach_div_ _foreach_mul_foreach_addcdiv_)rrRrSrTrUr"rXrYr r#rrrrVgrouped_tensorsgrouped_params_grouped_grads_grouped_exp_avgs_grouped_exp_infs_grouped_state_steps__grouped_params grouped_gradsgrouped_exp_avgsgrouped_exp_infsgrouped_state_stepsbias_correctionsr step_sizer1)rXrr r2_multi_tensor_adamax,s                   r)single_tensor_fnFr$c Cstjstdd|Dstd|durt||dd\}}|r*tjr*td|r4tjs4t}nt }||||||| | | | |||| |ddS) zrFunctional API that performs adamax algorithm computation. See :class:`~torch.optim.Adamax` for details. css|] }t|tjVqdSr^)r'r>r)rtr1r1r2rs  zadamax..zPAPI has changed, `state_steps` argument must contain a list of singleton tensorsNF) use_fusedz6torch.jit.script not supported with foreach optimizers) r"rXrYr r#rrrVr) r>rlrmrrMr jit is_scriptingrr)rrRrSrTrUr$rrrrVr"rXrYr r#rfuncr1r1r2rs>  )NFFFF)typingrrrrrr>r optimizerr r r r r rrrrrrrrr__all__r__doc__r@rbrrrr1r1r1r2s @    0      J      p