o ߗi>@svUddlmZmZmZmZddlZddlmmZ ddlm Z ddl m Z gZ eeed<ejjGdddZdS))DictListOptionalTupleN)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: )rrbeta1beta2rz%optimizer got an empty parameter listr)r ValueErrordefaultsrrtorchjitannotaterrstrstatelen param_group) selfrrrrrrrrr%g/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/torch/distributed/optim/functional_adamax.py__init__s. $z_FunctionalAdamax.__init__ gradientsc Cs|jd}g}g}g}g}g}t|t|kr*tddt|ddt|d}t|jd|D]]\} } | dur|t| O}|| || | |jvrwi|j| <|j| } td| d<tj | tj d | d <tj | tj d | d <|j| } || d || d || dq4t ,t j ||||||jd |jd |jd|jd|jd|j|j|d WddS1swYdS)NrzEthe gradients passed in does not equal to the size of the parameters!zParams length: z. zGradients length: Fr step) memory_formatexp_avgexp_infrrrrr)rrrrrrr has_complex)r#r"rzipr is_complexappendr!tensor zeros_likepreserve_formatno_gradFadamaxrrr) r$r(rparams_with_gradgradsexp_avgsexp_infs state_stepsr-paramgradientr!r%r%r&r)Csh          "z_FunctionalAdamax.stepN)r r r r FFF) __name__ __module__ __qualname__rrfloatrboolr'rr)r%r%r%r&r s4  )r )typingrrrrrtorch.optim._functionaloptim _functionalr5r,torch.distributed.optim._deprecation_warningrrr __annotations__rscriptr r%r%r%r&s