3jsSSKrSSKJr SSKJr SSKrSSKrSSKJ r J r SSK J r SSK JrJr \"SS \55r"S S \ \ 5rg) N) dataclass)Literal) ConfigMixinregister_to_config)SchedulerMixin) BaseOutput deprecatec\rSrSr%\R \S'Sr\R S-\S'Sr\R S-\S'Sr \ S-\S'Sr g)HeliosSchedulerOutput prev_sampleN model_outputs last_sample this_order) __name__ __module__ __qualname____firstlineno__torch FloatTensor__annotations__rrrint__static_attributes__r`/home/wildlama/miniconda3/lib/python3.13/site-packages/diffusers/schedulers/scheduling_helios.pyr r sF""".2M5$$t+2,0K""T)0!Jd !rr c%\rSrSr/rSr\SSS/SQSSS S S S S /S S SSS4S\S\S\S\ S\S\ S\ S\S\ S\ S\ S\ \S\ S\ S\ S\ S \ S!4"S"jj5rS#rS$r\S%5r\S&5rSWS'\4S(jjrS)rSXS*\S+\S -S,\ \R.-S-\ S -S.\ S -S/\ 4 S0jjrS.\S1\S2\R24S3jrS4rS5rSYS6jrS7rSZS8\R>S9\\R>-S:\R>S;\R@S -S1\R>S -S<\R>S -S=\ S>\!\"-4S?jjr#S@r$S S SA.S8\R2S:\R2S1\R2S>\R24SBjjr%S S S S SC.S8\R2S:\R2SD\S1\R2S<\R2S>\R24 SEjjr&S S S S S SF.SG\R2SH\R2SI\R2SD\SJ\R2S1\R2S>\R24SKjjr'S[S8\R2S9\\R2-S:\R2S=\ SL\ SM\ SJ\R2S1\R2S<\R2SN\SO\SP\SQ\R2S>\!\"-4SRjjr(S\S8\R>S9\\R>-S:\R>S;\R@S -S=\ S>\!\"-4 SSjjr)STr*SUr+SVr,g )]HeliosScheduler#i?)rUUUUUU?gUUUUUU?r!r$Fflow_predictionrTbh2Nunipc exponentialnum_train_timestepsshiftstages stage_rangegamma thresholdingprediction_type solver_order predict_x0 solver_typelower_order_finaldisable_correctorsolver_puse_flow_sigmasscheduler_typeuse_dynamic_shiftingtime_shift_type)r(linearc0Ul0Ul0Ul0Ul0Ul0UlUR 5 URSR5Ul URSR5Ul XPl U S;a0U S;aURSS9 O[U SUR35eXlS/U-UlS/U-UlSUlXlXlSUlSUlSUlg)Nr)bh1r&)midpointheunlogrhor&)r2z is not implemented for )timestep_ratiostimesteps_per_stagesigmas_per_stage start_sigmas end_sigmasori_start_sigmasinit_sigmas_for_each_stagesigmasitem sigma_min sigma_maxr-rNotImplementedError __class__r1r timestep_listlower_order_numsr4r5r _step_index _begin_index)selfr)r*r+r,r-r.r/r0r1r2r3r4r5r6r7r8r9s r__init__HeliosScheduler.__init__'s, "#% " " '')R--/Q,,. n ,<<''E':)[M9QRVR`R`Qa*bcc$"Vl2"Vl2 !!2  rcURRnURRn[R"SSU- US-5nSU- n[R "X$-SUS- U--- 5SSR 5n[R"U5nXA-R5nSUl SUl XPl URS5Ulg)z, initialize the global timesteps and sigmas r!r"Nr<cpu)configr)r*nplinspaceflipcopyr from_numpyclonerPrQ timestepstorH)rRr)r*alphasrHr^s r init_sigmasHeliosScheduler.init_sigmasZs#kk== !!Q$7 79Lq9PQv1 V/C+CDEcrJOOQ!!&)188:  "ii& rcUR5 /nURRnURRnURRn[ U5GH n[ XEU-5n[US5n[ XES-U-5n[Xs5nURUR5nXs:aURUR5OSn XRU'US:waKSU- n URRn S[R"SSU - -5SU - -U -- U -n SU - nURX- 5 XR U'XR"U'GM [%U5n [ U5HLnUS:XaSnO[%USU5U - nXRS- :XaSnO[%USUS-5U - nX4UR&U'MN [ U5GH nUR&Un[UR([ USU-5S5nUR([[ USU-5US- 5n[*R,"UUUS-5n[/U[0R25(aUSSO[0R4"USS5UR6U'[*R,"SSUS-5n[0R4"USS5UR8U'GM g) z# Init the timesteps for each stage rr!gNg?ir<g+?)rarWr+r)r,rangermaxminrHrIrFr-mathsqrtappendrDrEsumrAr^rXrY isinstancerTensorr\rBrC)rRstage_distancer+training_stepsr,i_s start_indice end_indice start_sigma end_sigma ori_sigmar-corrected_sigma tot_distance start_ratio end_ratiotimestep_ratio timestep_max timestep_minr^ stage_sigmass rrG*HeliosScheduler.init_sigmas_for_each_stagels ##88kk-- =C{/.@AL|Q/L[q1NBCJZ8J++l388:K:D:U J/446[^I)4 ! !# &ax O  ))#$ !q5y/(Ba)m(TW`(`#aen"n/1  ! !+"9 :%0  c "#,OOC '!,>* =Cax! !.#"67,F qj . yq 9:\I )4(@D  %!=C!11#6Nt~~c.2Cn2T.UVX[\L>>#c.2Cn2T.UWehiWi*jkL L,QR@RSI",Y "E"E #25K[K[\efigi\jKk  $ $S );;ua!1CDL).)9)9,s:K)LD ! !# &!rcUR$)zW The index counter for current timestep. It will increase 1 after each scheduler step. )rPrRs r step_indexHeliosScheduler.step_indexs rcUR$)za The index for the first timestep. It should be set from pipeline with `set_begin_index` method. rQrs r begin_indexHeliosScheduler.begin_indexs    rrcXlg)z Sets the begin index for the scheduler. This function should be run from pipeline before the inference. Args: begin_index (`int`): The begin index for the scheduler. Nr)rRrs rset_begin_indexHeliosScheduler.set_begin_indexs (rc2XRR-$NrWr))rRsigmas r _sigma_to_tHeliosScheduler._sigma_to_ts{{6666rnum_inference_steps stage_indexdevicerHmuis_amplify_first_chunkcURRS:XaU(a US-S-nOUS-nXlUR5 URRS:XaUc[ R "SSURR- US-5SSR[ R5nURRS:waDURR(aeURURRSU5nX@RR-R5n[R"U5nOUR Un[ R "USR#5USR#5U5nUR$Un [ R "U SR#5U SR#5U5n [R"U 5n[R"U5R'US9Ul[R*"U[R,"S5/5R'US9UlSUlUR35 URRS:XaKUR(SSUl[R*"UR.SS UR.SS/5UlURR(aURRS:XdeURUSUR.5UlURRS:Xa,UR.SSURR-UlgUR UR55UR.SSUR UR75UR UR55- --Ulgg) z1 Setting the timesteps and sigmas for each stage dmdrr!Nr<r"rr)rWr7rrar+rXrYr)astypefloat32r*r8 time_shiftr[rr\rBrIrCr_r^catzerosrHrPreset_scheduler_historyrfre) rRrrrrHrrr^stage_timestepsr|ratioss r set_timestepsHeliosScheduler.set_timestepss* ;; % % .%&9A&=&A#&9A&=##6   ;;   "~QDKK,K,K(KM`cdMdefigijqqJJ;;$$+#{{????!__T[[->->VLF++"A"AAGGII%%f-F"66{CO "'')#((*#I  00=L[[a!5!5!7b9I9N9N9PRefF%%f-F)))477v7FiiQ 89<[U[R5(d[U[R5(a [ S5eUR cSUlUR[R5nUc8Uc5URUR nURUR S-nX6U- U--nURUR5nU=RS- slU(dU4$[US9$)Nz:sigma and sigma_next must both be None or both be not NonezPassing integer indices (e.g. from `enumerate(timesteps)`) as timesteps to `EulerDiscreteScheduler.step()` is not supported. Make sure to pass one of the `scheduler.timesteps` as a timestep.rr!)r) rkrr IntTensor LongTensor ValueErrorrrPr_rrHdtyper ) rRrrrrrrrrs r step_eulerHeliosScheduler.step_euler7s :#56t8tt6 =Z/8S))h88h(8(899 K ?? " D 5==) =Z/KK0ET__q%89JU 2lBB "nn\%7%78  A> !$==rcURR(aSU- n[R"USS9nX#4$SUS-S-S-- nX-nX#4$)Nr!g:0yE>)rfr?)rWr6rclamp)rRralpha_tsigma_ts r_sigma_to_alpha_sigma_t'HeliosScheduler._sigma_to_alpha_sigma_tis\ ;; & &%iGkk%T2G E1HqLS01GoGrrrcH[U5S:aUSOURSS5nUc [U5S:aUSnO [S5eUb [SSS5 S nUcS nURUR nUR U5upUR(aURRS :Xa X)U-- U- n OURRS :XaUn OURRS :Xa X-X-- n OhURRS:Xa+U(aURUR n OUn X)U-- n O#[SURRS35eURR(aURU 5n U $URRS :XaU$URRS :Xa X(U-- U - n U $URRS :Xa X-X--n U $[SURRS35e)a Convert the model output to the corresponding type the UniPC algorithm needs. Args: model_output (`torch.Tensor`): The direct output from the learned diffusion model. timestep (`int`): The current discrete timestep in the diffusion chain. sample (`torch.Tensor`): A current instance of a sample created by the diffusion process. Returns: `torch.Tensor`: The converted model output. rrNr!/missing `sample` as a required keyword argumentr^1.0.0zPassing `timesteps` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`FTepsilonr v_predictionr%zprediction_type given as zj must be one of `epsilon`, `sample`, `v_prediction`, or `flow_prediction` for the UniPCMultistepScheduler.zW must be one of `epsilon`, `sample`, or `v_prediction` for the UniPCMultistepScheduler.) rpoprr rHrrr1rWr/r._threshold_sample) rRrrrargskwargsrflagrrx0_predrs rconvert_model_output$HeliosScheduler.convert_model_outputss."$i!m47J1M >4y1}a !RSS   Z   =DKK0E77> ??{{**i7!l$::gE,,8&,,>!*W-CC,,0AA"kk$//:G#G \#99 / 0K0K/LM\\ {{''009N{{**i7##,,8!l$::gE,,>!073CC / 0K0K/LMGGr)rorderrrrc [U5S:aUSOURSS5nUc [U5S:aUSnO [S5eUc [U5S:aUSnO [S5eUb [SSS 5 URn UR S n U S n Un UR (a(UR RXU 5Rn U $Uc8Uc5URURS-URURpOXTpURU5unnURU5unn[R"U5[R"U5- n[R"U5[R"U5- nUU- nURn/n/n[SU5HnURU- nU US-*nURURU5unn[R"U5[R"U5- nUU- U- nUR!U5 UR!UU - U- 5 M UR!S 5 [R""UUS 9n/n/n UR$(aU*OUn![R&"U!5n"U"U!- S- n#Sn$UR(R*S :XaU!n%O;UR(R*S:Xa[R&"U!5n%O [-5e[SUS-5HVnUR![R."UUS- 55 U R!U#U$-U%- 5 U$US--n$U#U!- SU$- - n#MX [R0"U5n[R""U US 9n [U5S:a[R0"USS9nUS:Xa"[R""S/U R2US9n&OW[R4R7USS 2SS 24U SS 5R9U5R9U R25n&OSnUR$(a:X- U -UU"-U -- n'Ub[R:"SW&U5n(OSn(U'UU%-U(-- n O:UU- U -UU"-U -- n'Ub[R:"SW&U5n(OSn(U'UU%-U(-- n U R9U R25n U $)aR One step for the UniP (B(h) version). Alternatively, `self.solver_p` is used if is specified. Args: model_output (`torch.Tensor`): The direct output from the learned diffusion model at the current timestep. prev_timestep (`int`): The previous discrete timestep in the diffusion chain. sample (`torch.Tensor`): A current instance of a sample created by the diffusion process. order (`int`): The order of UniP at this timestep (corresponds to the *p* in UniPC-p). Returns: `torch.Tensor`: The sample tensor at the previous timestep. r prev_timestepNr!rr.missing `order` as a required keyword argumentrzPassing `prev_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`r<r"rr=r&dimrrrk,bkc...->bc...)rrrr rrNr5steprrHrrrlogrrdritensorr1expm1rWr2rLpowstackrlinalgsolver_einsum))rRrrrrrrrrmodel_output_lists0m0xx_trsigma_s0ralpha_s0lambda_t lambda_s0hrrksD1sisimialpha_sisigma_si lambda_sirkRbhhh_phi_1h_phi_k factorial_iB_hrhos_px_t_pred_ress) rmultistep_uni_p_bh_update)HeliosScheduler.multistep_uni_p_bh_updates}6$'t9q=QfjjRV6W >4y1}a !RSS =4y1}Q !QRR  $ ^  !..    # r "  ==--$$\q9EECJ  %- $ DOOa,? @$++dooB^X *X77@!99(C(99W% '(::IIh'%))H*== y q%A1$B"QU8,B!%!=!=dkk"o!N Hh (+eii.AAIi'1,B JJrN JJR2~ &! 3ll3v.  ??aR++b/B," ;; " "e +C [[ $ $ -++b/C%' 'q%!)$A HHUYYsAE* + HHW{*S0 1 1q5 KlQ_4G % KKN LL6 * s8a<++cq)Czse1776J++Acrc3B3hK3B@CCFKNNqwwWC ??%)Gg,=,BBD <<(963G311CX%)Gg,=,BBD <<(963G311CffQWWo r)r this_sampler sigma_beforerthis_model_outputrr r c [U5S:aUSOURSS5n Uc [U5S:aUSnO [S5eUc [U5S:aUSnO [S5eUc [U5S:aUSnO [S 5eU b [SS S 5 URn U S n Un Un UnUc9Uc6UR UR UR UR S- nnOXennURU5unnURU5unn[R"U5[R"U5- n[R"U5[R"U5- nUU- nURn/n/n[SU5HnUR US-- nU US-*nURUR U5unn[R"U5[R"U5- nUU- U- nURU5 URUU - U- 5 M URS 5 [R"UUS9n/n /n!UR(aU*OUn"[R"U"5n#U#U"- S- n$Sn%UR R"S:XaU"n&O;UR R"S:Xa[R"U"5n&O [%5e[SUS-5HVnU R[R&"UUS- 55 U!RU$U%-U&- 5 U%US--n%U$U"- SU%- - n$MX [R("U 5n [R"U!US9n![U5S:a[R("USS9nOSnUS:Xa"[R"S/U R*US9n'OH[R,R/U U!5R1U5R1U R*5n'UR(aKUU- U -UU#-U -- n(Ub[R2"SU'SS U5n)OSn)X- n*U(UU&-U)U'S U*---- n OJUU- U -UU#-U -- n(Ub[R2"SU'SS U5n)OSn)X- n*U(UU&-U)U'S U*---- n U R1U R*5n U $)aI One step for the UniC (B(h) version). Args: this_model_output (`torch.Tensor`): The model outputs at `x_t`. this_timestep (`int`): The current timestep `t`. last_sample (`torch.Tensor`): The generated sample before the last predictor `x_{t-1}`. this_sample (`torch.Tensor`): The generated sample after the last predictor `x_{t}`. order (`int`): The `p` of UniC-p at this step. The effective order of accuracy should be `order + 1`. Returns: `torch.Tensor`: The corrected sample tensor at the current timestep. r this_timestepNr!z4missing `last_sample` as a required keyword argumentrz4missing `this_sample` as a required keyword argumentr#rrzPassing `this_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`r<r"rr=r&rrrr)rrrr rrHrrrrrrdrirr1rrWr2rLrrrrrr_r)+rRrrr rr rrrrrrrrmodel_trrrrrrrrrrrrrrrrrrrrrrrrrhos_crcorr_resD1_ts+ rmultistep_uni_c_bh_update)HeliosScheduler.multistep_uni_c_bh_updateIs<$'t9q=QfjjRV6W  4y1}"1g  !WXX  4y1}"1g  !WXX =4y1}Q !QRR  $ ^  !.. r " #  EM $ DOO ^XGX %XG77@!99(C(99W% '(::IIh'%))H*== y ##q%AAE*B"QU8,B!%!=!=dkk"o!N Hh (+eii.AAIi'1,B JJrN JJR2~ &! 3ll3v.  ??aR++b/B," ;; " "e +C [[ $ $ -++b/C%' 'q%!)$A HHUYYsAE* + HHW{*S0 1 1q5 KlQ_4G % KKN LL6 * s8a<++cq)CC A:\\3%qwwvFF\\''1-008;;AGGDF ??X%)Gg,=,BBD <<(96#2;L>7JT__7\]J11Jj$*?*?!*CD"""!44%//! 5   &$$t{{'?'??%%*%  !    ! 0@0@$//R R$#'((   rcURRS:XaURUUUUUS9$URRS:XaURUUUUS9$[e)Neuler)rrrrrr')rrrr)rWr7rrrL)rRrrrrrs rrHeliosScheduler.step>sy ;; % % 0??)!#' # [[ ' '7 2??)!' # & %rc$S/URR-UlS/URR-UlSUlURR UlURR UlSUlSUlSUl g)Nr) rWr0rrNrOr4r5rrPrQrs rr'HeliosScheduler.reset_scheduler_historyXsw"Vdkk&>&>>"Vdkk&>&>> !!%!>!> ,,  rc.URR$rrrs r__len__HeliosScheduler.__len__bs{{...r)rQrPr4rEr-rrOrrrFr1rKrJrHrCr5rDrrNrAr^rB)r)NNNNFr)NNNNNT) NNTNNNNNNNNN)NNNT)-rrrr _compatiblesrrrfloatlistboolstrrrrSrarGpropertyrrrrrrrrlrrrrrr Generatorr tuplerrrr rrrrr&rrrrrr#s L E$(0"0 "&')#' $%%*''/>%+++/>!! /> ??T) />   4' />%%,/>/>  &/>d  $" NllN N || N Nh $"#'DllD D  D || DLLD DT%)$(%)"L <<L\\ L \\ L  LllL||L Lb(,# ""%)"#'"#'"(,d lld  $d  d  d  d d lld ||d LLd d !d d d   &d T/3$(,0 &''&%+++&!! & ??T) &  &  &&4!/rr)rg dataclassesrtypingrnumpyrXrconfiguration_utilsrrschedulers.scheduling_utilsrutilsr r r rrrrr6sJ ! A8) "J" "@ /nk@ /r