tjY#ddlmZddlZddlmZddlmZmZmZmZm Z m Z m Z m Z m Z mZmZmZmZmZddgZGddeZd d e d ed e d ed ze_ d%deedeedeedeededzdedzdedzdededzdedededededededdf"d Zd!Zdeedeedeedeededzdedzdededededededededdfd"Zdeedeedeedeededzdedzdededededededededdfd#Zdeedeedeedeededzdedzdeezdedededededededdfd$ZdS)&)castN)Tensor)_default_to_fused_or_foreach_device_dtype_check_for_fused_differentiable_doc _foreach_doc_get_scalar_dtype _get_value _maximize_doc _params_doc _to_scalar_use_grad_for_differentiable _view_as_real DeviceDict OptimizerParamsTAdagradadagradceZdZ ddddddedeezd ed ed ed ed edzdedededzddffdZfdZddZ dZ e ddZ xZ S)r{Gz?r绽|=NF)maximizedifferentiablefusedparamslrlr_decay weight_decayinitial_accumulator_valueepsforeachrrrreturnc t|tr'|dkrtdd|kstd|d|kstd|d|kstd|d|kstd|d|kstd||||||||| | d } t || | r0| rt d |rt d d |_d |_|j D]} | d D]} |j | }| dr0tj dt| d| jn!tjdt|d<tj| rt#||n|}tj| |tj|d<dS)NrzTensor lr must be 1-elementzInvalid learning rate: zInvalid lr_decay value: zInvalid weight_decay value: z)Invalid initial_accumulator_value value: zInvalid epsilon value: ) rrr!rr r"rrrz)`fused` does not support `differentiable`z0`fused` and `foreach` cannot be `True` together.Trris_fuseddtypedevicer*step memory_formatsum) isinstancernumel ValueErrorsuper__init__ RuntimeError"_need_device_dtype_check_for_fused_step_supports_amp_scaling param_groupsstatetorchzerosr r+tensor is_complexcomplex full_likepreserve_format)selfrrrrr r!r"rrrdefaultsgrouppr: init_value __class__s Q/var/www/html/CA-Chatbot/venv/lib/python3.11/site-packages/torch/optim/adagrad.pyr5zAdagrad.__init__sk b& ! ! } tj t| dt|  | d<=dSdS) Nr"rFrrrrr-r'r)r,)r4 __setstate__r9 setdefaultr:getlenr; is_tensorfloatr=r r+listvalues) rBr:rrDrEp_statestep_val state_valuesstep_is_tensorsrGs rHrKzAdagrad.__setstate__ds U###&  E   Y - - -   Z / / /   -u 5 5 5$$Wd33E8_  *..B//w<<1$$U_WV_-M-M$$WV_55H!> O $"3U"C"C"C#$8 #\(:K:M:MNNNFO DJ--//00 l++q0 eo OF #7 7  !  !L!F)$$,=u,M,M,M&     rIc~|jD]4}|dD])}|j|}|d*5dS)z6Calls tensor.share_memory_() on the state sum tensors.rr0N)r9r: share_memory_)rBrDrEr:s rH share_memoryzAdagrad.share_memorysZ& - -E8_ - - 1 e **,,,, - - -rIcd\}}|dD]}|j|dr't|ddrt|d|_||jjz}|t j|z}||||j|j|} t| dkr|drt||dr0t j dt|d |j n!t j d t | d <|jd} t j|rt| | n| } t j|| t j| d<|| d|| d ||fS)N)FFrrr7TFrr&r'r)r%r,r-r r.r0)gradgetattrrr7 is_sparser;r>appendr:rNr<r r+r=rCr?r@rA) rBrDparams_with_gradgrads state_sums state_stepshas_sparse_grad has_complexrEr:r rFs rH _init_groupzAdagrad._init_groups'3$x( 2( 2Av!>Dg8''D 2!444>CD;16#33u/222  ''*** QV$$$ 1 u::??W~95a888!> J "3U7^"L"L"L#$8 #\#5F5H5HIII&M15 31- !+A..7 9;TUUU6 $)?:U5J$$$E%L!!%,///""5=111 ++rIcd}|5tj5|}dddn #1swxYwY|jD]}g}g}g}g}||||||\}} t |||||d|d|d|d||d|d|d| |d t |d dt |d d |S) zPerform a single optimization step. Args: closure (Callable, optional): A closure that reevaluates the model and returns the loss. Nrrrr!r"rrr grad_scale found_inf) rrrr!rdr"rrrerrhri)r; enable_gradr9rfrr]) rBclosurelossrDr`rarbrcrdres rHr-z Adagrad.steps`  "$$ ! !wyy ! ! ! ! ! ! ! ! ! ! ! ! ! ! !&  E-/ "$E')J(*K+/+;+;' K,, (O[  ;">2z*%L /i(z*$%56'Gn"4t<<!$ T::!     & s /33)rrrrrN)r#NN)__name__ __module__ __qualname__rrPrboolr5rKrZrfrr- __classcell__)rGs@rHrrsW"+,#E$!EEEE FNE E  E $) EEEEEd{E EEEEEEN!!!!!F----,,,,,,\"***"!*****rIa[Implements Adagrad algorithm. .. math:: \begin{aligned} &\rule{110mm}{0.4pt} \\ &\textbf{input} : \gamma \text{ (lr)}, \: \theta_0 \text{ (params)}, \: f(\theta) \text{ (objective)}, \: \lambda \text{ (weight decay)}, \\ &\hspace{12mm} \tau \text{ (initial accumulator value)}, \: \eta\text{ (lr decay)}\\ &\textbf{initialize} : state\_sum_0 \leftarrow \tau \\[-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} \tilde{\gamma} \leftarrow \gamma / (1 +(t-1) \eta) \\ &\hspace{5mm} \textbf{if} \: \lambda \neq 0 \\ &\hspace{10mm} g_t \leftarrow g_t + \lambda \theta_{t-1} \\ &\hspace{5mm}state\_sum_t \leftarrow state\_sum_{t-1} + g^2_t \\ &\hspace{5mm}\theta_t \leftarrow \theta_{t-1}- \tilde{\gamma} \frac{g_t}{\sqrt{state\_sum_t}+\epsilon} \\ &\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 `Adaptive Subgradient Methods for Online Learning and Stochastic Optimization`_. z Args: a lr (float, Tensor, optional): learning rate (default: 1e-2) lr_decay (float, optional): learning rate decay (default: 0) weight_decay (float, optional): weight decay (L2 penalty) (default: 0) initial_accumulator_value (float, optional): initial value of the sum of squares of gradients (default: 0) eps (float, optional): term added to the denominator to improve numerical stability (default: 1e-10) z a fused (bool, optional): whether the fused implementation (CPU and CUDA only) is used. Currently, `torch.float64`, `torch.float32`, `torch.float16`, and `torch.bfloat16` are supported. (default: None). Please note that the fused implementation does not support sparse or complex gradients. .. _Adaptive Subgradient Methods for Online Learning and Stochastic Optimization: http://jmlr.org/papers/v12/duchi11a.html Frrarbrcrrhrirdr"rrerrrr!rr#c 4td|Dstd||t|| d\}}|d}|d}|r-tjrtd|r-tjrtd|r&tjst }n/|r&tjst}nt}|||||| | | |||| | ||dS) ztFunctional API that performs Adagrad algorithm computation. See :class:`~torch.optim.Adagrad` for details. c3JK|]}t|tjVdSrm)r1r;r).0ts rH zadagrad..6s.@@qz!U\**@@@@@@rIzPAPI has changed, `state_steps` argument must contain a list of singleton tensorsNF) use_fusedz6torch.jit.script not supported with foreach optimizersz4torch.jit.script not supported with fused optimizers rrrr!rdrrrerhri) allr6rr;jit is_scripting_fused_adagrad_multi_tensor_adagrad_single_tensor_adagrad)rrarbrcrrhrirdr"rrerrrr!r_funcs rHrrsn2 @@K@@@ @ @  ^    }1 Ne   7 }U59))++USTTT S''))SQRRR &UY++--& &//11&$%D  ! '%rIcV|}tj|||Srm)sizer;sparse_coo_tensor)r\ grad_indicesrRrs rH _make_sparsergs$ 99;;D  "< > >>rIc >||tdtjst |}t ||||dD]F\}}}}|dz }t |}| s|n| }|dkr-|jrtd| ||}|d|dz |zzz }|jr| }| }| }| t|||d||}|  | }| t||||z | Xtj|}|r view_as_realaddcmul_sqrtaddcdiv_view_as_complex)rrarbrcrhrirrrr!rdrrreparamr\ state_sumstep_tr-clrr grad_valuesstd std_valuesr>s rHrrls"!6KLLL 9 ! ! # # ^^*-z;t+++*=*=&tY ! &!!#.tt$ 1  ~ "Q88E866DAX--. > ===??D==??L,,..K NN<lKOOA.s59 9 #DN9 9 9 9 9 9 rITryg?cpu)r+rrcHg|]} dt|dz zzz S)r)r )rur-rrs rH z)_multi_tensor_adagrad.. sD   >BRC1 4((1,88 9   rIr)rrNrr"_group_tensors_by_device_and_dtyperRrrQranyrrr; _foreach_negcompiler is_compilingis_cpu _foreach_add_r= _foreach_add_foreach_addcmul_ _foreach_sqrt _foreach_mul_ _foreach_mul_foreach_addcdiv_)rrarbrcrhrirrrr!rdrrregrouped_tensorlistsdevice_params_ device_grads_device_state_sums_device_state_steps_r device_params device_gradsdevice_state_sumsdevice_state_stepsdevice_has_sparse_grad minus_clrr numerators ` ` rHr~r~sH"DBCCC!6KLLL 6{{a BB#F  K0 & & ( ( M?M?  T&\>:: DL-88  f/ABB!$v,0CDD!0" S9 9 '39 9 9 6 6  "  "!")! $!-'%#       J -7H I I I  < -l;;L ~**,, 71CA1F1M 7  "ELU$C$C$C3        2A 6 6 6 1   #L-|TTTTT$1 -|         FX     1<UVWWWW!"344 C%%% 1      i 8 8 8$II*<CCI  y#>>>>[M?M?rIc>|sdS| s| rtd| rtd| |j|ini}| |j|ini}t|tr!t |jdkr |j|ind}t j||||g}|D]t\\}}\\}}}}}ttt|}ttt|}ttt|}ttt|}d\}}|+| || |d}|+| || |d}|&||vr"| |d||<||}tj |dtj|||||||| | || |&tj||gt!|zvdS) Nz5`fused` does not support sparse grad or complex paramz 9%%B&b&11Wc")nn6M6MBSW  B  K0O    ++       T&\>:: DL-88  f/ABB!$v,0CDD.8++  ! / : : f4 @@!!   -88 V$ ??     6#8#8 ee6eEEGFOB .222     %(&   '  "%5$6=O9P9P$P   S++rI)NNNFNFF)typingrr;r optimizerrrrr r r r r rrrrrr__all__r__doc__rQrqrPrrrr~r}r&rIrHrs  $ i KKKKKiKKK^4    5.p $#" GG LG <GV Gf G $; G  G}GGD[GGG G !G"#G$ %G&'G( )GGGGT??? A= LA= <A=V A=f A=  A= } A= A=A=A= A=A=A=A=A=  !A=A=A=A=Hl? Ll? <l?V l?f l?  l? } l? l?l?l? l?l?l?l?l?  !l?l?l?l?^S LS <SV Sf S  S } S SSS SSSSS  !SSSSSSrI