o H h+@sjUddlmZddlZddlmmZddlmZddlm Z gZ e e e d<ejjGdddZdS))OptionalN)Tensor)2_scripted_functional_optimizer_deprecation_warning__all__c@sheZdZ       ddeededeeefd ed ed ed ed efddZdee efddZ dS)_FunctionalAdamaxMbP?g?g+?:0yE>Fparamslrbetaseps weight_decayforeachmaximize_allow_empty_param_listc Cstddd|kstd|d|kstd|d|dkr'dks1ntd|dd|d kr=dksGntd |d d|ksRtd ||||d|d |d |_||_||_tjttj tt tj ffi|_ t |dkr|std d|i|_ dS)N) stacklevelr zInvalid learning rate: zInvalid epsilon value: rg?z#Invalid beta parameter at index 0: z#Invalid beta parameter at index 1: zInvalid weight_decay value: )r rbeta1beta2rz%optimizer got an empty parameter listr )r ValueErrordefaultsrrtorchjitannotatedictrstrstatelen param_group) selfr r r rrrrrr#]/var/www/vscode/kcb/lib/python3.10/site-packages/torch/distributed/optim/functional_adamax.py__init__s. $z_FunctionalAdamax.__init__ gradientsc Cs|jd}g}g}g}g}g}t|t|kr*tddt|ddt|d}t|jd|D]]\} } | dur|t| O}|| || | |jvrwi|j| <|j| } td| d<tj | tj d | d <tj | tj d | d <|j| } || d || d || dq4t ,t j ||||||jd |jd |jd|jd|jd|j|j|d WddS1swYdS)Nr zEthe gradients passed in does not equal to the size of the parameters!zParams length: z. zGradients length: Fr step) memory_formatexp_avgexp_infrrrr r)rrrr rrr has_complex)r!r rzipr is_complexappendrtensor zeros_likepreserve_formatno_gradFadamaxrrr) r"r&r params_with_gradgradsexp_avgsexp_infs state_stepsr+paramgradientrr#r#r$r'Csh          "z_FunctionalAdamax.stepN)rrr r FFF) __name__ __module__ __qualname__listrfloattupleboolr%rr'r#r#r#r$rs4  )r)typingrrtorch.optim._functionaloptim _functionalr3r,torch.distributed.optim._deprecation_warningrrr?r__annotations__rscriptrr#r#r#r$s