!Ei6: SrSSKJrJr SSKrSSKJs Jr SSK J r /SQr "SS5r "SS 5r "S S 5r\"S \ S 9rSS\S\S\S\S\S-S\4 SjjrSS\S\S\4Sjjrg)z=Spectral Normalization from https://arxiv.org/abs/1802.05957.)AnyTypeVarN)Module) SpectralNorm SpectralNormLoadStateDictPreHookSpectralNormStateDictHook spectral_normremove_spectral_normc 4\rSrSr%Sr\\S'\\S'\\S'\\S'\\S'SS\S\S\S\S S 4 S jjr S\ RS \ R4S jr S \ S\S \ R4SjrS \ S S 4SjrS \ S\S S 4SjrSr\S \ S\S\S\S\S S4 Sj5rSrg )r_versionnamedimn_power_iterationsepsweightreturnNc\XlX0lUS::a[SU35eX lX@lg)NrzGExpected n_power_iterations to be positive, but got n_power_iterations=)rr ValueErrorrr)selfrrrrs Y/home/mande/repo/quber/.venv/lib/python3.13/site-packages/torch/nn/utils/spectral_norm.py__init__SpectralNorm.__init__#s@   "**<)=? #5c"UnURS:waVUR"UR/[UR55Vs/sHo3UR:wdMUPM snQ76nURS5nUR US5$s snf)Nr)rpermuterangesizereshape)rr weight_matdheights rreshape_weight_to_matrix%SpectralNorm.reshape_weight_to_matrix4s 88q=#++',Z^^-='>P'>!txx-A'>PJ#!!&"--Qs B B moduledo_power_iterationc @[XRS-5n[XRS-5n[XRS-5nURU5nU(a[R"5 [ UR 5H|n[R"[R"UR5U5SURUS9n[R"[R"Xe5SURUS9nM~ UR S:a:UR[RS9nUR[RS9nSSS5 [R"U[R"Xe55nX8- nU$!,(df  N?=f)N_orig_u_vr)rrout) memory_format)getattrrr%torchno_gradrrF normalizemvtrclonecontiguous_formatdot) rr'r(ruvr"_sigmas rcompute_weightSpectralNorm.compute_weight>s0<W!45 FII, - FII, -226: t667A 3aA EHHZ$;VWXA8**Q.e.E.EFAe.E.EFA! !UXXj45 !!s 6C F Fc[R"5 URUSS9nSSS5 [XR5 [XRS-5 [XRS-5 [XRS-5 UR UR[R RWR555 g!,(df  N=f)NFr(r+r,r*) r0r1r=delattrrregister_parameternn Parameterdetach)rr'rs rremoveSpectralNorm.removets ]]_((E(JF " D() D() G+,!!$))UXX-?-? -PQ _s C C"inputsc `[UURURXRS95 g)Nr@)setattrrr=training)rr'rHs r__call__SpectralNorm.__call__}s)  II   ??  K rc n[RRUR5R U5R 5UR5UR S5/5RS5nURU[R"U[R"X55- 5$)Nr ) r0linalg multi_dotr5mmpinverse unsqueezesqueezemul_r8r4)rr"r9 target_sigmar:s r_solve_v_and_rescale!SpectralNorm._solve_v_and_rescales LL " " \\^  z * 3 3 5z||~q{{ST~ V '!* vvlUYYq%((:2I%JJKKrcURR5H8n[U[5(dMURU:XdM,[ SU35e [XX45nUR UnUc[SUS35e[U[RRR5(a [S5e[R"5 URU5nUR5up[R "UR#U 5R%SS5SUR&S9n [R "UR#U 5R%SS5SUR&S9n SSS5 [)XR5 UR+URS-U5 [-XRUR.5 UR1URS -W 5 UR1URS -W 5 UR3U5 UR5[7U55 UR9[;U55 U$!,(df  N=f) Nz>Cannot register two spectral_norm hooks on the same parameter z/`SpectralNorm` cannot be applied as parameter `z ` is NonezThe module passed to `SpectralNorm` can't have uninitialized parameters. Make sure to run the dummy forward before applying spectral normalizationrr )rrr*r+r,)_forward_pre_hooksvalues isinstancerr RuntimeError _parametersrr0rC parameterUninitializedParameterr1r%r r2r3 new_emptynormal_rrArBrJdataregister_bufferregister_forward_pre_hook_register_state_dict_hookr"_register_load_state_dict_pre_hookr) r'rrrrhookfnrr"hwr9r:s rapplySpectralNorm.applys--446D$ --$))t2C"TUYTZ[7 $C =##D) >A$yQ  fehh00GG H H\  ]]_44V# UH nTU-T;v M g7frr{).0s state_dict weight_keys r .sT>S Q*4>Ss)r*r+r,F)r*r+Tr*r+r,) rigetrallappendr0r1popmeanr%rW)rrprefixlocal_metadatastrict missing_keysunexpected_keys error_msgsriversionhas_missing_keyssuffixkey weight_origrr<r"r9r:rs ` @rrL)SpectralNormLoadStateDictPreHook.__call__sFWW $$_b9== GGj $  ?gk'')JT>STTTj0 $ - 6)j('+$v$++C0 .  (g)=> # 3$-33588E zD01++J5A01 :,-!)*(!s A D55 ErrNrorprqrrrrLrzr{rrrrs)2 )2rrc(\rSrSrSSjrSSjrSrg)rNcXlgrrrs rr"SpectralNormStateDictHook.__init__rrcSU;a0US'URRS-nXTS;a[SU35eURRUSU'g)Nr rz-Unexpected key in metadata['spectral_norm']: )rirr]r)rr'rrrrs rrL"SpectralNormStateDictHook.__call__s] . 0.0N? +ggllZ' 1 1!NseTU U/3ww/?/?',rrrrr{rrrrs @rrT_module)boundr'rrrrrcUca[U[RR[RR[RR 45(aSnOSn[ RXX$U5 U$)aApply spectral normalization to a parameter in the given module. .. math:: \mathbf{W}_{SN} = \dfrac{\mathbf{W}}{\sigma(\mathbf{W})}, \sigma(\mathbf{W}) = \max_{\mathbf{h}: \mathbf{h} \ne 0} \dfrac{\|\mathbf{W} \mathbf{h}\|_2}{\|\mathbf{h}\|_2} Spectral normalization stabilizes the training of discriminators (critics) in Generative Adversarial Networks (GANs) by rescaling the weight tensor with spectral norm :math:`\sigma` of the weight matrix calculated using power iteration method. If the dimension of the weight tensor is greater than 2, it is reshaped to 2D in power iteration method to get spectral norm. This is implemented via a hook that calculates spectral norm and rescales weight before every :meth:`~Module.forward` call. See `Spectral Normalization for Generative Adversarial Networks`_ . .. _`Spectral Normalization for Generative Adversarial Networks`: https://arxiv.org/abs/1802.05957 Args: module (nn.Module): containing module name (str, optional): name of weight parameter n_power_iterations (int, optional): number of power iterations to calculate spectral norm eps (float, optional): epsilon for numerical stability in calculating norms dim (int, optional): dimension corresponding to number of outputs, the default is ``0``, except for modules that are instances of ConvTranspose{1,2,3}d, when it is ``1`` Returns: The original module with the spectral norm hook .. note:: This function has been reimplemented as :func:`torch.nn.utils.parametrizations.spectral_norm` using the new parametrization functionality in :func:`torch.nn.utils.parametrize.register_parametrization`. Please use the newer version. This function will be deprecated in a future version of PyTorch. Example:: >>> m = spectral_norm(nn.Linear(20, 40)) >>> m Linear(in_features=20, out_features=40, bias=True) >>> m.weight_u.size() torch.Size([40]) r r)r\r0rCConvTranspose1dConvTranspose2dConvTranspose3drrl)r'rrrrs rr r sjp {  ((((((   CCv%7cB MrcURR5HLup#[U[5(dMURU:XdM.UR U5 URU O [ SUSU35eURR5HEup#[U[5(dMURRU:XdM8URU O URR5HFup#[U[5(dMURRU:XdM8URU U$ U$)zRemove the spectral normalization reparameterization from a module. Args: module (Module): containing module name (str, optional): name of weight parameter Example: >>> m = spectral_norm(nn.Linear(40, 10)) >>> remove_spectral_norm(m) zspectral_norm of 'z' not found in ) rZitemsr\rrrFr_state_dict_hooksrri_load_state_dict_pre_hooksr)r'rkrhs rr r Ss,,224 dL ) )dii4.? KK ))!,  5 -dV?6(KLL++113 d5 6 6477<<4;O((+ 4 44::< d< = =$'',,RVBV11!4  M = Mr)rr rnN)r)__doc__typingrrr0torch.nn.functionalrC functionalr2torch.nn.modulesr__all__rrrrrursrvr r r{rrrsC # eeT6262v @ @ :V ,  F F FF  F t F  FRHr