a d) @sddlZddlmZddlmZmZmZmZmZmZddl m Z ddl m Z m Z ddgZGd ddeZd d jeeed e_de ee ee ee ee eeeeeeed ddZe ee ee ee eeeeeeed ddZe ee ee ee eeeeeeed ddZdS)N)Tensor) Optimizer_use_grad_for_differentiable_default_to_fused_or_foreach_differentiable_doc _foreach_doc _maximize_doc)"_group_tensors_by_device_and_dtype)ListOptionalAdadeltaadadeltacsVeZdZddddeeeedfd d Zfd d Zd dZedddZ Z S)r ??ư>rNF)maximizedifferentiableforeachrrc sd|kstd|d|kr*dks:ntd|d|ksPtd|d|ksftd|t|||||||d} t|| dS)NgzInvalid learning rate: {}rzInvalid rho value: {}zInvalid epsilon value: {}zInvalid weight_decay value: {})lrrhoeps weight_decayrrr) ValueErrorformatdictsuper__init__) selfparamsrrrrrrrdefaults __class__]/var/www/html/stable-diffusion-webui/venv/lib/python3.9/site-packages/torch/optim/adadelta.pyr s$  zAdadelta.__init__cs@t||jD](}|dd|dd|ddqdS)NrrFr)r __setstate__ param_groups setdefault)rstategroupr"r$r%r&-s     zAdadelta.__setstate__cCs|dD]}|jdurq|||jjr2td||j|j|}t|dkrd|d<tj|tjd|d<tj|tjd|d<||d||d|dd7<qdS) Nr z*Adadelta does not support sparse gradientsrstep)Z memory_format square_avg acc_deltar) gradappendZ is_sparse RuntimeErrorr)lentorchZ zeros_likeZpreserve_format)rr*params_with_gradgrads square_avgs acc_deltaspr)r$r$r% _init_group4s&        zAdadelta._init_groupcCsd}|dur:t|}Wdn1s00Y|jD]}g}g}g}g}|d|d|d|d|d|d|df\}} } } } } }||||||t|||||| | | | | |d q@|S) zPerforms a single optimization step. Args: closure (Callable, optional): A closure that reevaluates the model and returns the loss. Nrrrrrrr)rrrrrrr)r2Z enable_gradr'r8r)rclosureZlossr*r3r4r5r6rrrrrrrr$r$r%r+Ns@ $  z Adadelta.step)rrrrN)N) __name__ __module__ __qualname__r boolrr&r8rr+ __classcell__r$r$r"r%r s"  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`_. a Args: params (iterable): iterable of parameters to optimize or dicts defining parameter groups rho (float, optional): coefficient used for computing a running average of squared gradients (default: 0.9) eps (float, optional): term added to the denominator to improve numerical stability (default: 1e-6) lr (float, optional): coefficient that scale delta before it is applied to the parameters (default: 1.0) weight_decay (float, optional): weight decay (L2 penalty) (default: 0) {foreach} {maximize} {differentiable} .. _ADADELTA\: An Adaptive Learning Rate Method: https://arxiv.org/abs/1212.5701 rF) r r4r5r6rrrrrrrc Csh|durt||dd\} }|r0tjr0td|rDtjsDt} nt} | |||||||| | |d dS)zvFunctional API that performs Adadelta algorithm computation. See :class:`~torch.optim.Adadelta` for details. NF)Z use_fusedz6torch.jit.script not supported with foreach optimizers)rrrrrr)rr2ZjitZ is_scriptingr0_multi_tensor_adadelta_single_tensor_adadelta) r r4r5r6rrrrrrr_funcr$r$r%rs&) r r4r5r6rrrrrrcCst||||D]\} } } } |s"| n| } |dkr>| j| |d} t| rft| } t| } t| } | |j| | d|d| |}| |}| r|}| || | |j||d|dt| rt |}| j || dqdS)Nralpharvalue) zipaddr2Z is_complexZ view_as_realZmul_Zaddcmul_Zsqrt_cloneZdiv_Zview_as_complexZadd_)r r4r5r6rrrrrrparamr.r,r-stddeltar$r$r%r@s(      r@) r r4r5r6rrrrrrcCs| r Jdt|dkrdSt||||g} | D]\} } } }|rNt| } |dkrftj| | |d} t| |tj| | | d|dt| |}t|t||}t|t ||t|| tj | || dt||tj|||d|dq4dS)Nz#_foreach ops don't support autogradrrCrrE) r1r valuesr2Z _foreach_negZ _foreach_addZ _foreach_mul_Z_foreach_addcmul_Z_foreach_sqrt_Z _foreach_div_Z _foreach_add_)r r4r5r6rrrrrrZgrouped_tensorsZ device_paramsZ device_gradsZdevice_square_avgsZdevice_acc_deltasrKZdeltasr$r$r%r?s(           r?)NF)r2rZ optimizerrrrrrr Ztorch.utils._foreach_utilsr typingr r __all__r r__doc__r=floatrr@r?r$r$r$r%s`   q7 0 )