+hLSSKrSSKJrJrJrJrJr SSKrSSK r SSK J r J r SSK JrJr SSKJr SSKJrJrJr \"5(aSSKrS S jr"S S \\ 5rg) N)CallableListOptionalTupleUnion) ConfigMixinregister_to_config) deprecateis_scipy_available) randn_tensor)KarrasDiffusionSchedulersSchedulerMixinSchedulerOutputc &US:XaSnOUS:XaSnO[SU35e/n[U5H<nXP- nUS-U- nUR[SU"U5U"U5- - U55 M> [R "U[R S9$)a Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [0,1]. Contains a function alpha_bar that takes an argument t and transforms it to the cumulative product of (1-beta) up to that part of the diffusion process. Args: num_diffusion_timesteps (`int`): the number of betas to produce. max_beta (`float`): the maximum beta to use; use values lower than 1 to prevent singularities. alpha_transform_type (`str`, *optional*, default to `cosine`): the type of noise schedule for alpha_bar. Choose from `cosine` or `exp` Returns: betas (`np.ndarray`): the betas used by the scheduler to step the model outputs cosinech[R"US-S- [R-S- 5S-$)NgMb?gT㥛 ?r)mathcospits b/home/james-whalen/.local/lib/python3.13/site-packages/diffusers/schedulers/scheduling_sasolver.py alpha_bar_fn)betas_for_alpha_bar..alpha_bar_fn<s-88QY%/$''9A=>!C Cexpc4[R"US-5$)Ng()rrrs rrrAs88AI& &rz"Unsupported alpha_transform_type: rdtype) ValueErrorrangeappendmintorchtensorfloat32)num_diffusion_timestepsmax_betaalpha_transform_typerbetasit1t2s rbetas_for_alpha_barr0#s.x' D  & '=>R=STUU E * +  (!e. . S\"- R0@@@(KL, <<U]] 33rc0\rSrSrSr\VVs/sHoR PM snnrSr\ SSSSSS S S SS S S SSS S S S S \ "S5*SSS4S\ S\ S\ S\ S\ \\R \\ 4S\ S\ S\ S\ \S\S\ S\ S\ S \S!\ \S"\ \S#\ \S$\ \S%\ \ S&\ S'\ \ S(\ S)\ 4.S*jj5r\S+5r\S,5rSXS-\ 4S.jjrSYS/\ S0\\ \R444S1jjrS2\R8S3\R84S4jrS5rS6rS7\R8S3\R84S8jr S7\R8S/\ S3\R84S9jr!SZS7\R8S/\ S:\ S;\ S3\R84 S<jjr"SS=.S>\R8S2\R8S3\R84S?jjr#S@r$SAr%SBr&SCr'S>\R8S2\R8SD\R8SE\ SF\R8S3\R84 SGjr(SH\R8SI\R8SJ\R8SK\R8SE\ SF\R8S3\R84SLjr)S[SMjr*SNr+S\S>\R8SO\ S2\R8SP\S3\\,\-44 SQjjr.S2\R8S3\R84SRjr/SS\R8SD\R8ST\R`S3\R84SUjr1SVr2SWr3gs snnf)]SASolverSchedulerOu  `SASolverScheduler` is a fast dedicated high-order solver for diffusion SDEs. This model inherits from [`SchedulerMixin`] and [`ConfigMixin`]. Check the superclass documentation for the generic methods the library implements for all schedulers such as loading and saving. Args: num_train_timesteps (`int`, defaults to 1000): The number of diffusion steps to train the model. beta_start (`float`, defaults to 0.0001): The starting `beta` value of inference. beta_end (`float`, defaults to 0.02): The final `beta` value. beta_schedule (`str`, defaults to `"linear"`): The beta schedule, a mapping from a beta range to a sequence of betas for stepping the model. Choose from `linear`, `scaled_linear`, or `squaredcos_cap_v2`. trained_betas (`np.ndarray`, *optional*): Pass an array of betas directly to the constructor to bypass `beta_start` and `beta_end`. predictor_order (`int`, defaults to 2): The predictor order which can be `1` or `2` or `3` or '4'. It is recommended to use `predictor_order=2` for guided sampling, and `predictor_order=3` for unconditional sampling. corrector_order (`int`, defaults to 2): The corrector order which can be `1` or `2` or `3` or '4'. It is recommended to use `corrector_order=2` for guided sampling, and `corrector_order=3` for unconditional sampling. prediction_type (`str`, defaults to `epsilon`, *optional*): Prediction type of the scheduler function; can be `epsilon` (predicts the noise of the diffusion process), `sample` (directly predicts the noisy sample`) or `v_prediction` (see section 2.4 of [Imagen Video](https://imagen.research.google/video/paper.pdf) paper). tau_func (`Callable`, *optional*): Stochasticity during the sampling. Default in init is `lambda t: 1 if t >= 200 and t <= 800 else 0`. SA-Solver will sample from vanilla diffusion ODE if tau_func is set to `lambda t: 0`. SA-Solver will sample from vanilla diffusion SDE if tau_func is set to `lambda t: 1`. For more details, please check https://huggingface.co/papers/2309.05019 thresholding (`bool`, defaults to `False`): Whether to use the "dynamic thresholding" method. This is unsuitable for latent-space diffusion models such as Stable Diffusion. dynamic_thresholding_ratio (`float`, defaults to 0.995): The ratio for the dynamic thresholding method. Valid only when `thresholding=True`. sample_max_value (`float`, defaults to 1.0): The threshold value for dynamic thresholding. Valid only when `thresholding=True` and `algorithm_type="dpmsolver++"`. algorithm_type (`str`, defaults to `data_prediction`): Algorithm type for the solver; can be `data_prediction` or `noise_prediction`. It is recommended to use `data_prediction` with `solver_order=2` for guided sampling like in Stable Diffusion. lower_order_final (`bool`, defaults to `True`): Whether to use lower-order solvers in the final steps. Default = True. use_karras_sigmas (`bool`, *optional*, defaults to `False`): Whether to use Karras sigmas for step sizes in the noise schedule during the sampling process. If `True`, the sigmas are determined according to a sequence of noise levels {σi}. use_exponential_sigmas (`bool`, *optional*, defaults to `False`): Whether to use exponential sigmas for step sizes in the noise schedule during the sampling process. use_beta_sigmas (`bool`, *optional*, defaults to `False`): Whether to use beta sigmas for step sizes in the noise schedule during the sampling process. Refer to [Beta Sampling is All You Need](https://huggingface.co/papers/2407.12173) for more information. lambda_min_clipped (`float`, defaults to `-inf`): Clipping threshold for the minimum value of `lambda(t)` for numerical stability. This is critical for the cosine (`squaredcos_cap_v2`) noise schedule. variance_type (`str`, *optional*): Set to "learned" or "learned_range" for diffusion models that predict variance. If set, the model's output contains the predicted Gaussian variance. timestep_spacing (`str`, defaults to `"linspace"`): The way the timesteps should be scaled. Refer to Table 2 of the [Common Diffusion Noise Schedules and Sample Steps are Flawed](https://huggingface.co/papers/2305.08891) for more information. steps_offset (`int`, defaults to 0): An offset added to the inference steps, as required by some model families. rig-C6?g{Gz?linearNrepsilonFgףp= ??data_predictionTinflinspacernum_train_timesteps beta_startbeta_end beta_schedule trained_betaspredictor_ordercorrector_orderprediction_typetau_func thresholdingdynamic_thresholding_ratiosample_max_valuealgorithm_typelower_order_finaluse_karras_sigmasuse_exponential_sigmasuse_beta_sigmasuse_flow_sigmas flow_shiftlambda_min_clipped variance_typetimestep_spacing steps_offsetcURR(a[5(d [S5e[ URRURR URR /5S:a [S5eUb)[R"U[RS9Ul OUS:Xa*[R"X#U[RS9Ul OkUS:Xa4[R"US-US-U[RS9S-Ul O1US :Xa[U5Ul O[US UR35eS UR- Ul[R""UR S S 9Ul[R&"UR$5Ul[R&"SUR$- 5Ul[R,"UR(5[R,"UR*5- UlSUR$- UR$- S-UlS UlU S;a[U S UR35eSUl[6R"S US- U[6RS9SSS2R95n[R:"U5UlS/[?XgS- 5-Ul S/[?XgS- 5-Ul!U c SUl"OXl"U S:HUl#S Ul$SUl%SUl&SUl'UR0RQS5Ulg)Nz:Make sure to install scipy if you want to use beta sigmas.rznOnly one of `config.use_beta_sigmas`, `config.use_exponential_sigmas`, `config.use_karras_sigmas` can be used.r r4 scaled_linear?rsquaredcos_cap_v2z is not implemented for r6rdim)r7noise_predictionc"US:aUS::aS$S$)Ni rrrs r,SASolverScheduler.__init__..s18Sa&Ga&Grr7cpu))configrJr ImportErrorsumrIrHr"r&r'r(r,r9r0NotImplementedError __class__alphascumprodalphas_cumprodsqrtalpha_tsigma_tloglambda_tsigmasinit_noise_sigmanum_inference_stepsnpcopy from_numpy timestepsmax timestep_list model_outputsrB predict_x0lower_order_nums last_sample _step_index _begin_indexto)selfr:r;r<r=r>r?r@rArBrCrDrErFrGrHrIrJrKrLrMrNrOrPrrs r__init__SASolverScheduler.__init__s6 ;; & &/A/C/CZ[ [  ++T[[-O-OQUQ\Q\QnQno pst tA   $m5==IDJ h & >QY^YfYfgDJ o -OcM'--    J1 1,-@ADJ%7OPTP^P^O_&`a aDJJ& #mmDKKQ?zz$"5"56 zz!d&9&9"9:  $,,/%))DLL2II D///43F3FF3N !$ !H H%(88PQUQ_Q_P`&ab b$( KK#6#:(a[R"SSURR - US-5n S U - n[R"URR@U-SURR@S- U--- 5SSR5nXpRR -R5n[R2"XwSS/5R[R45nO[RB"U[R "S[EU55U5nSUR(S- UR(S- S -n [R2"X{//5R[R45n[RF"U5Ul$[RF"U5RKU[RS 9Ul&[EU5Ul'S/[QURRRURRTS- 5-Ul+SUl,SUl-SUl.SUl/URHRKS 5Ul$gs sn fs sn fs sn f)au Sets the discrete timesteps used for the diffusion chain (to be run before inference). Args: num_inference_steps (`int`): The number of diffusion steps used when generating samples with a pre-trained model. device (`str` or `torch.device`, *optional*): The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. rr9rNrXleadingtrailingzY is not supported. Please make sure to choose one of 'linspace', 'leading' or 'trailing'.rS) in_sigmasrnr6)rr!r^)0r& searchsortedfliprkr_rMr:numpyitemrOror9roundrpastypeint64arangerPr"arrayrfrjrH_convert_to_karras _sigma_to_t concatenater(rI_convert_to_exponentialrJ_convert_to_betarKrLinterplenrqrlr{rrrnrsr?r@rurwrxryrz) r|rnr clipped_idx last_timesteprr step_ratiorl log_sigmassigmard sigma_lasts r set_timestepsSASolverScheduler.set_timestepss((DMMA3)GIgIgh ++99KGNNPVVX  ;; ' ': 5 A}q02E2IJPPRSWUWSWXY\Z\]bbdkklnltltu [[ ) )Y 6&+BCJ1&9A&=>KRRTUYWYUYZ[^\^_ddfmmnpnvnvwI 11 1I [[ ) )Z 788;NNJ -ZK@FFHMMOVVWYW_W_`I NI;;//01JK A 3 33t7J7JJsRSVVF^ ;; ( (WWV_))+F,,v,gFSY!ZSY%"2"25"ESY!Z[aacI^^VBC[$9:AA"**MF [[ / /WWV_))+F11F1lFSY!ZSY%"2"25"ESY!Z[I^^VBC[$9:AA"**MF [[ ( (WWV_))+F**V*eFSY!ZSY%"2"25"ESY!Z[I^^VBC[$9:AA"**MF [[ ( ([[A (G(G$GI\_`I`aF6\FWWT[[33f<T[[E[E[^_E_ciDi@ijklomopuuwF++"A"AAGGII^^VBC[$9:AA"**MFYYy"))As6{*CVLFt221559L9LQ9OOTWWJ^^V\$:;BB2::NF&&v. )))477vU[[7Y#&y>   ++T[[-H-H1-L MN!"  kknnU+ I"[ "[ "[s.] 8]4]samplereturncXURnURtp4nU[R[R4;aUR 5nUR X4[R"U5-5nUR5n[R"X`RRSS9n[R"USURRS9nURS5n[R"X*U5U- nUR "X4/UQ76nUR!U5nU$)aC "Dynamic thresholding: At each sampling step we set s to a certain percentile absolute pixel value in xt0 (the prediction of x_0 at timestep t), and if s > 1, then we threshold xt0 to the range [-s, s] and then divide by s. Dynamic thresholding pushes saturated pixels (those near -1 and 1) inwards, thereby actively preventing pixels from saturation at each step. We find that dynamic thresholding results in significantly better photorealism as well as better image-text alignment, especially when using very large guidance weights." https://huggingface.co/papers/2205.11487 rrU)r%rs)r!shaper&r(float64floatreshaperoprodabsquantiler_rDclamprE unsqueezer{)r|rr! batch_sizechannelsremaining_dims abs_sampless r_threshold_sample#SASolverScheduler._threshold_sampleWs 06 - ~  6 6\\^F rww~7N,NOZZ\ NN:{{'M'MST U KK 1$++66  KKNVR+a/ F~F5! rc[R"[R"US55nX2SS2[R4- n[R"US:SS9R SS9R URSS- S9nUS-nX%nX&nXs- Xx- - n [R "U SS5n SU - U-X--n U RUR5n U $)Ng|=r)axisr)rsr) rorjmaximumnewaxiscumsumargmaxcliprr) r|rr log_sigmadistslow_idxhigh_idxlowhighwrs rrSASolverScheduler._sigma_to_tysFF2::eU34 q"**}55))UaZq188a8@EE*JZJZ[\J]`aJaEbQ;!#_ , GGAq! Ug  , IIekk "rcvURR(a SU- nUnX#4$SUS-S-S-- nX-nX#4$)NrrrS)r_rK)r|rrhris r_sigma_to_alpha_sigma_t)SASolverScheduler._sigma_to_alpha_sigma_tsQ ;; & &%iGG E1HqLS01GoGrrc[URS5(aURRnOSn[URS5(aURRnOSnUbUOUSR 5nUbUOUSR 5nSn[ R "SSU5nUSU- -nUSU- -nXXx- --U-n U $)z6Constructs the noise schedule of Karras et al. (2022). sigma_minN sigma_maxrXrg@r)hasattrr_rrrror9) r|rrnrrrhoramp min_inv_rho max_inv_rhorls rr$SASolverScheduler._convert_to_karrass 4;; , , --II 4;; , , --II!*!6IIbM The algorithm and model type are decoupled. You can use either data_prediction or noise_prediction for both noise prediction and data prediction models. Args: model_output (`torch.Tensor`): The direct output from the learned diffusion model. sample (`torch.Tensor`): A current instance of a sample created by the diffusion process. Returns: `torch.Tensor`: The converted model output. rrNr/missing `sample` as a required keyword argumentrr1.0.0zPassing `timesteps` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`)r7r5)learned learned_ranger v_predictionflow_predictionzprediction_type given as zd must be one of `epsilon`, `sample`, `v_prediction`, or `flow_prediction` for the SASolverScheduler.)rWzQ must be one of `epsilon`, `sample`, or `v_prediction` for the SASolverScheduler.)rpopr"r rlrrr_rFrArNrCrrhri) r|rrargskwargsrrrhrix0_predr5s rconvert_model_output&SASolverScheduler.convert_model_outputs:"$i!m47J1M >4y1}a !RSS   Z   DOO,77> ;; % %)< <{{**i7;;,,0LL#/2A2#6L!l$::gE,,8&,,>!*W-CC,,0AA++doo6 \#99 / 0K0K/LMVV {{''009N[[ ' '+? ?{{**i7;;,,0LL*1bqb51G*G,,8!l$::gE,,>!073CC / 0K0K/LMAA {{''#'<<#94<<;Q!g$55@009!g$55@N/@rcUS;dS5eUS:Xa3[R"U*5[R"X2- 5S- -$US:Xa<[R"U*5US-[R"X2- 5-US-- -$US:XaN[R"U*5US-SU--S-[R"X2- 5-US-SU--S-- -$US:Xa`[R"U*5US-SUS---SU--S-[R"X2- 5-US-SUS---SU--S-- -$g) zT Calculate the integral of exp(-x) * x^order dx from interval_start to interval_end rrrr)order is only supported for 0, 1, 2 and 3rrrrNr&r)r|orderinterval_start interval_ends r%get_coefficients_exponential_negative7SASolverScheduler.get_coefficients_exponential_negativeJsz $Q&QQ$ A:99l]+uyy9V/WZ[/[\ \ aZ99l]+!#uyy1N'OOS_bcScd aZ99l]+"Q%77!;uyyIf?gg?Q%559; aZ99l]+"Q):%::Q=OORSS))L9:;?Qq%881|;KKaOQ rcUS;dS5eSUS--U-nSUS--U-nUS:Xa<[R"U5S[R"XV- *5- -SUS--- $US:XaH[R"U5US- US- [R"XV- *5-- -SUS--S-- $US:XaZ[R"U5US-SU-- S-US-SU-- S-[R"XV- *5-- -SUS--S-- $US:Xal[R"U5US-SUS--- SU--S- US-SUS--- SU--S- [R"XV- *5-- -SUS--S-- $g ) z\ Calculate the integral of exp(x(1+tau^2)) * x^order dx from interval_start to interval_end rrrrrrrNr)r|rrrtauinterval_end_covinterval_start_covs r%get_coefficients_exponential_positive7SASolverScheduler.get_coefficients_exponential_positivebs $Q&QQ$QJ,6#q&jN: A: *+q599?O?d=e3f/fgklortuoukuv aZ *+%))A-=M=b;c1dde QJ1$ & aZ *+%q(1/?+??!C)1,q3E/EEIii"2"G HIJJ QJ1$ & aZ *+%q(1/?/B+BBQIYEYY\]])1,q3Eq3H/HH1OaKaadeeii"2"G HIJJ QJ1$ & rc US;deU[U5S- :XdeUS:XaS//$US:Xa@SUSUS- - US*USUS- - /SUSUS- - US*USUS- - //$US:XaUSUS- USUS- -nUSUS- USUS- -nUSUS- USUS- -nSU- US*US- U- USUS-U- /SU- US*US- U- USUS-U- /SU- US*US- U- USUS-U- //$US:XGaUSUS- USUS- -USUS- -nUSUS- USUS- -USUS- -nUSUS- USUS- -USUS- -nUSUS- USUS- -USUS- -nSU- US*US- US- U- USUS-USUS--USUS--U- US*US-US-U- /SU- US*US- US- U- USUS-USUS--USUS--U- US*US-US-U- /SU- US*US- US- U- USUS-USUS--USUS--U- US*US-US-U- /SU- US*US- US- U- USUS-USUS--USUS--U- US*US-US-U- //$g)z2 Calculate the coefficient of lagrange polynomial rrrrrN)r)r|r lambda_list denominator1 denominator2 denominator3 denominator4s rlagrange_polynomial_coefficient1SASolverScheduler.lagrange_polynomial_coefficientsa  $$$K(1,,,, A:C5L aZQ+a.89 ^O{1~ A'FG Q+a.89 ^O{1~ A'FG  aZ'N[^; AQ\]^Q_@_`L'N[^; AQ\]^Q_@_`L'N[^; AQ\]^Q_@_`L $!!n_{1~5EN[^3lB  $!!n_{1~5EN[^3lB  $!!n_{1~5EN[^3lB "aZQ+a.0q>KN24q>KN24  Q+a.0q>KN24q>KN24  Q+a.0q>KN24q>KN24  Q+a.0q>KN24q>KN24  $!!n_{1~5 AF,V#AQ7%a.;q>9:%a.;q>9:# # "!n_{1~5 AF,V  $!!n_{1~5 AF,V#AQ7%a.;q>9:%a.;q>9:# # "!n_{1~5 AF,V  $!!n_{1~5 AF,V#AQ7%a.;q>9:%a.;q>9:# # "!n_{1~5 AF,V  $!!n_{1~5 AF,V#AQ7%a.;q>9:%a.;q>9:# # "!n_{1~5 AF,V E-- +rc US;deU[U5:XdS5e/nURUS- U5n[U5HnSn [U5H]n UR(a&XUU UR US- U - X#U5-- n M:XUU UR US- U - X#5-- n M_ UR U 5 M [U5U:XdS5eU$)N)rrrrz4the length of lambda list must be equal to the orderrrz3the length of coefficients does not match the order)rrr#rvrrr$) r|rrrrr coefficientslagrange_coefficientr- coefficientjs rget_coefficients_fn%SASolverScheduler.get_coefficients_fns  $$$K((`*``( #CCEAI{[uAK5\??#:1#=@j@j A ~SA$K #:1#=@j@j A ~A$K "    ,< E)`+``)rnoiserrc  [U5S:aUSOURSS5nUc [U5S:aUSnO [S5eUc [U5S:aUSnO [S5eUc [U5S:aUSnO [S 5eUc [U5S :aUS nO [S 5eUb [SS S 5 URn UR UR S-UR UR pURU 5upURU 5up[R"U 5[R"U 5- n[R"U 5[R"U 5- n[R"U5nX- n/n[U5HqnUR U- nURUR U5unn[R"U5[R"U5- nURU5 Ms URXOUUU5nUnUR(GaTUS:XGaMUR UR S- nURU5unn[R"U5[R"U5- nUS==S[R"SUS--U-5-US-S- USUS---S- [R"SUS--U*-5-SUS--S-- - -UU- - - ss'US==S[R"SUS--U-5-US-S- USUS---S- [R"SUS--U*-5-SUS--S-- - -UU- - -ss'[U5HqnUR(a>USUS--U -[R"US-*U-5-UU-U US-*-- nMRUSUS--*U -UU-U US-*-- nMs UR(a=U [R "S[R"SUS--U-5- 5-U-nO8XZ-[R "[R"SU-5S- 5-U-nUR(a,[R"US-*U-5X- -U-U-U-nO X- U-U-U-nUR#UR$5nU$)a One step for the SA-Predictor. Args: model_output (`torch.Tensor`): The direct output from the learned diffusion model at the current timestep. prev_timestep (`int`): The previous discrete timestep in the diffusion chain. sample (`torch.Tensor`): A current instance of a sample created by the diffusion process. order (`int`): The order of SA-Predictor at this timestep. Returns: `torch.Tensor`: The sample tensor at the previous timestep. r prev_timestepNrrrz.missing `noise` as a required keyword argumentr.missing `order` as a required keyword argumentr,missing `tau` as a required keyword argumentrzPassing `prev_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`r6rrr"r rurlrrr&rj zeros_liker#r$r rvrrgr{r!) r|rrrrrrrrmodel_output_listrisigma_s0rhalpha_s0rk lambda_s0 gradient_parthrr-sialpha_sisigma_si lambda_sigradient_coefficientsx temp_sigma temp_alpha_s temp_sigma_s temp_lambda_s noise_partx_ts r!stochastic_adams_bashforth_update3SASolverScheduler.stochastic_adams_bashforth_updates!6$'t9q=QfjjRV6W >4y1}a !RSS =4y1}Q !QRR =4y1}Q !QRR ;4y1}1g !OPP  $ ^  !.. KK!+ , KK ( 77@!99(C99W% '(::IIh'%))H*== ((0   uA1$B!%!=!=dkk"o!N Hh (+eii.AAI   y )  !% 8 88U`be f  ??? "[[1)<= -1-I-I*-U* l % , 7%))L:Q Q %a(iiS!Vx 789!tax1CF #3a#7%))QaZUVTVDW:X#X^_beghbh^hmn]n"ooq!=02( &a(iiS!Vx 789!tax1CF #3a#7%))QaZUVTVDW:X#X^_beghbh^hmn]n"ooq!=02(uAaZii#q& H 456,A./(!a%1 2 1sAv:!8;PQR;S!SVgjknojohpVq!qq  ?? 5::a%))BaK!O2L.L#MMPUUJEIIa!e4Dq4H)IIEQJ ??))c1fIM*g.@AAE UXbbC%*]:ZGCffQWWo rthis_model_outputrx last_noise this_samplec  [U5S:aUSOURSS5n Uc [U5S:aUSnO [S5eUc [U5S:aUSnO [S5eUc [U5S:aUSnO [S 5eUc [U5S :aUS nO [S 5eUc [U5S :aUS nO [S 5eU b [SSS5 URn UR UR UR UR S- pURU 5upURU 5up[R"U 5[R"U 5- n[R"U5[R"U 5- n[R"U5nUU- n/n[U5HqnUR U- nURUR U5unn[R"U5[R"U5- nURU5 Ms X/-nURUUUUU5nUnUR(aUS:XaUS==S[R"SUS--U-5-US- USUS---S- [R"SUS--U*-5-SUS--S-U-- - -- ss'US==S[R"SUS--U-5-US- USUS---S- [R"SUS--U*-5-SUS--S-U-- - --ss'[U5HqnUR(a>USUS--U -[R"US-*U-5-UU-UUS-*-- nMRUSUS--*U -UU-UUS-*-- nMs UR(a=U [R "S[R"SUS--U-5- 5-U-nO8Xk-[R "[R"SU-5S- 5-U-nUR(a,[R"US-*U-5X- -U-U-U-nO X- U-U-U-nUR#UR$5nU$)a One step for the SA-Corrector. Args: this_model_output (`torch.Tensor`): The model outputs at `x_t`. this_timestep (`int`): The current timestep `t`. last_sample (`torch.Tensor`): The generated sample before the last predictor `x_{t-1}`. this_sample (`torch.Tensor`): The generated sample after the last predictor `x_{t}`. order (`int`): The order of SA-Corrector at this step. Returns: `torch.Tensor`: The corrected sample tensor at the current timestep. r this_timestepNrz4missing `last_sample` as a required keyword argumentrz3missing `last_noise` as a required keyword argumentrz4missing `this_sample` as a required keyword argumentrrrrzPassing `this_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`r6rr)r|r+rxr,r-rrrrr/rrirrhrrkrrrrr-rrrr model_prev_listr!r"r'r(s rstochastic_adams_moulton_update1SASolverScheduler.stochastic_adams_moulton_updates>$'t9q=QfjjRV6W  4y1}"1g  !WXX  4y1}!!W  !VWW  4y1}"1g  !WXX =4y1}Q !QRR ;4y1}1g !OPP  $ ^  !.. KK ( KK!+ , 77@!99(C99W% '(::IIh'%))H*== ((5 y  uA1$B!%!=!=dkk"o!N Hh (+eii.AAI   y )  ,.AA $ 8 8 8U`be f  ?? &a(iiS!Vx 7891uQaZ 01 4uyy!c1f*RSQSAT7U U[\_bde_e[ejkZknoZoppr( &a(iiS!Vx 7891uQaZ 01 4uyy!c1f*RSQSAT7U U[\_bde_e[ejkZknoZoppr( uAaZii#q& H 456,A./&Ah/ 0 1sAv:!8;PQR;S!SVehilmhmfnVo!oo  ?? 5::a%))BaK!O2L.L#MMPZZJEIIa!e4Dq4H)IIJVJ ??))c1fIM*g.@AAE UXbbC%*]:ZGCffQWWo rcUc URnX!:HR5n[U5S:Xa[UR5S- nU$[U5S:aUSR5nU$USR5nU$)Nrr)rrnonzerorr)r|rschedule_timestepsindex_candidatesrs rindex_for_timestep$SASolverScheduler.index_for_timesteps  %!% .:CCE  A %T^^,q0J ! "Q &)!,113J*!,113JrcURc[[U[R5(a%UR UR R 5nURU5UlgURUlg)z6 Initialize the step_index counter for the scheduler. N) r isinstancer&Tensorr{rrrr8ryrz)r|rs r_init_step_index"SASolverScheduler._init_step_index(sZ    #(ELL11#;;t~~'<'<=#66x@D #00D rr return_dictc URc [S5eURcURU5 URS:=(a URSLnUR XS9nU(aPUR URS5nURUURURUURUS9n[[URRURRS- 5S- 5HAn UR U S-UR U 'URU S-URU 'MC XpR S'X RS'[#UR$UUR&UR(S9n URR*(a[-URR[/UR05UR- 5n [-URR[/UR05UR- S-5n O,URRn URRn [-XR2S-5Ul[-XR2S -5Ul UR4S:deURS:deX0lXl UR URS5nUR7UUU UR4US 9n UR2[URRURRS- 5:aU=R2S- slU=R8S- slU(dU 4$[;U S 9$) ae Predict the sample from the previous timestep by reversing the SDE. This function propagates the sample with the SA-Solver. Args: model_output (`torch.Tensor`): The direct output from learned diffusion model. timestep (`int`): The current discrete timestep in the diffusion chain. sample (`torch.Tensor`): A current instance of a sample created by the diffusion process. generator (`torch.Generator`, *optional*): A random number generator. return_dict (`bool`): Whether or not to return a [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`. Returns: [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`: If return_dict is `True`, [`~schedulers.scheduling_utils.SchedulerOutput`] is returned, otherwise a tuple is returned where the first element is the sample tensor. NzaNumber of inference steps is 'None', you need to run 'set_timesteps' after creating the schedulerrrrX)r+rxr,r-rrr) generatorrr!r)rrrrr) prev_sample)rnr"rr=rxrrBrtr2r,this_corrector_orderr#rsr_r?r@rur rrr!rGr%rrrrwthis_predictor_orderr)ryr)r|rrrrAr? use_correctormodel_output_convert current_taur-rrDrCrBs rstepSASolverScheduler.step4s<  # # +s  ?? "  ! !( +!+L0@0@0L #888U --(:(:2(>?K99"6 ,,??"// :Fs4;;66 8S8SVW8WX[\\]A$($6$6q1u$=D  q !$($6$6q1u$=D  q !^"62!)2   &&$$   ;; ( (#&t{{'B'BCDWZ^ZiZiDi#j #&t{{'B'BCDWZ^ZiZiDilmDm#n #';;#>#> #';;#>#> $'(<>S>SVW>W$X!$'(<>S>SVW>W$X!((1,,,((1,,,!mmD$6$6r$:; <<-++ =   3t{{'B'BDKKD_D_bcDc#d d  ! !Q & ! A> !;77rcU$)z Ensures interchangeability with schedulers that need to scale the denoising model input depending on the current timestep. Args: sample (`torch.Tensor`): The input sample. Returns: `torch.Tensor`: A scaled input sample. r[)r|rrrs rscale_model_input#SASolverScheduler.scale_model_inputs  roriginal_samplesrrcURRURS9UlURRURS9nURUR5nXCS-nUR 5n[ UR 5[ UR 5:a?URS5n[ UR 5[ UR 5:aM?SXC- S-nUR 5n[ UR 5[ UR 5:a?URS5n[ UR 5[ UR 5:aM?XQ-Xb--nU$)N)rr rSrXr)rfr{rr!flattenrrr)r|rMrrrrfsqrt_alpha_prodsqrt_one_minus_alpha_prod noisy_sampless r add_noiseSASolverScheduler.add_noisesW#1144>$=qFF$=>L E$("%BF  ('+",1"%/"&,116*/*/&)%*5\M'+ *1S, S,S, S,  S,  bjj$u+&= >? S,S,S,S,8$S,S,%*S, S,S, S, $D>!S,"!)#S,$"$%S,&"$'S,(UO)S,*"+S,, }-S,./S,01S,S,j  !!(3(O,O,U3PUP\P\K\E]O,d D0 ELLRWR^R^4TW\a\h\h.dgrpsQ$ 99 A1,XX !)4Xv/ v/r