+
    ƒ°ñj¢/  ã                   ó@  € ^ RI t ^ RIt^ RIt^ RIHt ^ RIt^ RIHtH	t	 ^ RI
Ht  ! R R]P                  4      tRR ltRR ltRR ltRR	 ltR
 tR t ! R R]P(                  P*                  4      tRR ltR tR tR tR t ! R R]P                  4      tRR ltR# )é    N)ÚrepeatÚ	rearrange)Úinstantiate_from_configc                   ór   a a€ ] tR t^t o. ROtR	V3R lV 3R llltV3R lR ltR
V3R lR lltRtVt	V ;t
# )ÚAlphaBlenderc                ó,   <€ V ^8„  d   QhRS[ RS[RS[/# )é   ÚalphaÚmerge_strategyÚrearrange_pattern)ÚfloatÚstr)ÚformatÚ__classdict__s   "€ÚKC:\Ai_Coding\apps\studio\ComfyUI\comfy\ldm\modules\diffusionmodules\util.pyÚ__annotate__ÚAlphaBlender.__annotate__   s.   ø€ ÷ Nñ NáðNñ ðNñ ñ	Nó    c                óô  <€ \         SV `  4        W n        W0n        W P                  9   g   Q R V P                   24       hV P                  R8X  d*   V P                  R\        P                  ! V.4      4       R# V P                  R8X  g   V P                  R8X  dG   V P                  R\        P                  P                  \        P                  ! V.4      4      4       R# \        RV P                   24      h)zmerge_strategy needs to be in ÚfixedÚ
mix_factorÚlearnedÚlearned_with_imageszunknown merge strategy N)ÚsuperÚ__init__r   r   Ú
strategiesÚregister_bufferÚtorchÚTensorÚregister_parameterÚnnÚ	ParameterÚ
ValueError)Úselfr
   r   r   Ú	__class__s   &&&&€r   r   ÚAlphaBlender.__init__   sÕ   ø€ ô 	‰ÑÔØ,ÔØ!2Ôð Ÿo™oÔ-ð	>à+¨D¯O©OÐ+<Ð=ó	>Ø-ð ×Ñ 'Ô)Ø× Ñ  ¬u¯|ª|¸U¸GÓ/DÖEà×Ñ 9Ô,Ø×"Ñ"Ð&;Ô;à×#Ñ#ØœeŸh™h×0Ñ0´·²¸u¸gÓ1FÓGöô Ð6°t×7JÑ7JÐ6KÐLÓMÐMr   c                óN   <€ V ^8„  d   QhRS[ P                  RS[ P                  /# )r	   Úimage_only_indicatorÚreturn©r   r   )r   r   s   "€r   r   r   1   s#   ø€ ÷ ñ ©e¯l©lð ÁuÇ|Á|ñ r   c                óÒ  € V P                   R 8X  d   V P                  P                  V4      pV# V P                   R8X  d2   \        P                  ! V P                  P                  V4      4      pV# V P                   R8X  dÞ   Vf;   \        \        P                  ! V P                  P                  V4      4      R4      pMˆ\        P                  ! VP                  4       \        P                  ! ^^VP                  R7      \        \        P                  ! V P                  P                  VP                  4      4      R4      4      p\        W0P                  4      pV# \        4       h)r   r   r   z... -> ... 1©Údevice)r   r   Útor   Úsigmoidr   ÚwhereÚboolÚonesr-   r   ÚNotImplementedError)r$   r(   r-   r
   s   &&& r   Ú	get_alphaÚAlphaBlender.get_alpha1   s  € à×Ñ 'Ô)ð —O‘O×&Ñ& vÓ.ˆEð& ˆð% × Ñ  IÔ-Ü—M’M $§/¡/×"4Ñ"4°VÓ"<Ó=ˆEð" ˆð × Ñ Ð$9Ô9Ø#Ò+Ü!¤%§-¢-°·±×0BÑ0BÀ6Ó0JÓ"KÈ^Ó\‘äŸšØ(×-Ñ-Ó/Ü—J’J˜q !Ð,@×,GÑ,GÔHÜœeŸmšm¨D¯O©O×,>Ñ,>Ð?S×?ZÑ?ZÓ,[Ó\Ð^lÓmó�ô
 ˜e×%;Ñ%;Ó<ˆEð
 ˆô &Ó'Ð'r   c                ó4   <€ V ^8„  d   QhRS[ P                  /# )r	   r)   r*   )r   r   s   "€r   r   r   K   s   ø€ ÷ ñ ñ
 
�‰ñr   c                óÜ   € V P                  W1P                  4      pVP                  VP                  4      V,          R V,
          P                  VP                  4      V,          ,           pV# )g      ð?)r4   r-   r.   Údtype)r$   Ú	x_spatialÚ
x_temporalr(   r
   Úxs   &&&&  r   ÚforwardÚAlphaBlender.forwardK   sZ   € ð —‘Ð3×5EÑ5EÓFˆà�H‰H�Y—_‘_Ó%¨	Õ1Ø�U�{×Ñ˜yŸ™Ó/°*Õ<õ=ð 	
ð ˆr   )r   r   )r   r   r   )r   zb t -> (b t) 1 1©N)Ú__name__Ú
__module__Ú__qualname__Ú__firstlineno__r   r   r4   r<   Ú__static_attributes__Ú__classdictcell__Ú__classcell__©r%   r   s   @@r   r   r      s/   ù‡ € Ú<€J÷Nõ N÷4ð ÷4÷ ò r   r   c                 óP  € V R 8X  d@   \         P                  ! VR,          VR,          V\         P                  R7      ^,          pV# V R8X  dÑ   \         P                  ! V^,           \         P                  R7      V,          V,           pV^V,           ,          \        P
                  ,          ^,          p\         P                  ! V4      P                  ^4      pWw^ ,          ,          p^VR,          VRR ,          ,
          p\         P                  ! V^ RR7      pV# V R8X  d   \        VR	 4      # V R
8X  d*   \         P                  ! W#V\         P                  R7      pV# V R8X  d1   \         P                  ! W#V\         P                  R7      R,          pV# \        RV  R24      h)Úlinearg      à?)r8   Úcosineºé   NNNç+‡ÙÎ÷ï?)ÚminÚmaxÚsquaredcos_cap_v2c                 ó�   € \         P                  ! V R ,           R,          \         P                  ,          ^,          4      ^,          # )çü©ñÒMb€?gTã¥›Ä ð?)ÚmathÚcosÚpi)Úts   &r   Ú<lambda>Ú$make_beta_schedule.<locals>.<lambda>m   s)   € ”d—h’h  E¥	¨UÕ2´T·W±WÕ<¸qÕ@ÓAÀQÖFr   Úsqrt_linearÚsqrtz
schedule 'z
' unknown.éÿÿÿÿ)r   ÚlinspaceÚfloat64ÚarangeÚnprT   rS   ÚpowÚclampÚbetas_for_alpha_barr#   )ÚscheduleÚ
n_timestepÚlinear_startÚ
linear_endÚcosine_sÚbetasÚ	timestepsÚalphass   &&&&&   r   Úmake_beta_schedulerj   Y   sl  € Ø�8Ôä—’˜|¨sÕ2°JÀ#Õ4EÀzÔY^×YfÑYfÔgÐklÕlð 	ð6 €Lð/ 
�XÔ	ä—’˜Z¨!�^´5·=±=ÔAÀJÕNÐQYÕYð 	ð ˜a (�lÕ+¬b¯e©eÕ3°aÕ7ˆÜ—’˜6Ó"×&Ñ& qÓ)ˆØ �)Õ#ˆØ�F˜2•J ¨¨ Õ,Õ,ˆÜ—’˜E q¨eÔ4ˆð €Lð 
Ð(Ô	(ä"ØÙFó
ð 	
ð
 
�]Ô	"Ü—’˜|¸Ì5Ï=É=ÔYˆð
 €Lð	 
�VÔ	Ü—’˜|¸Ì5Ï=É=ÔYÐ]`Õ`ˆð €Lô ˜: h Z¨zÐ:Ó;Ð;r   c           	      ó”  € V R 8X  d3   W!,          p\         P                  ! \        \        ^ W$4      4      4      pMdV R8X  dO   \         P                  ! ^ \         P
                  ! VR,          4      V4      ^,          P                  \        4      pM\        RV  R24      hV^,           pV'       d   \        P                  ! RV 24       V# )ÚuniformÚquadgš™™™™™é?z/There is no ddim discretization method called "Ú"z%Selected timesteps for ddim sampler: )r^   ÚasarrayÚlistÚranger[   rY   ÚastypeÚintr3   ÚloggingÚinfo)Úddim_discr_methodÚnum_ddim_timestepsÚnum_ddpm_timestepsÚverboseÚcÚddim_timestepsÚ	steps_outs   &&&&   r   Úmake_ddim_timestepsr}   y   sª   € Ø˜IÔ%ØÕ4ˆÜŸš¤D¬¨qÐ2DÓ)HÓ$IÓJ‰Ø	˜fÔ	$ÜŸ;š; q¬"¯'ª'Ð2DÀrÕ2IÓ*JÐL^Ó_ÐdeÕe×mÑmÔnqÓr‰ä!Ð$SÐTeÐSfÐfgÐ"hÓiÐið  Õ"€IßÜ�ŠÐ<¸Y¸KÐHÔIØÐr   c                 óž  € W,          p\         P                  ! V ^ ,          .WRR ,          P                  4       ,           4      pV\         P                  ! ^V,
          ^V,
          ,          ^WE,          ,
          ,          4      ,          pV'       d9   \        P
                  ! RV RV 24       \        P
                  ! RV RV 24       WdV3# )r   Nz'Selected alphas for ddim sampler: a_t: z; a_(t-1): z&For the chosen value of eta, which is zB, this results in the following sigma_t schedule for ddim sampler rZ   )r^   ro   ÚtolistrY   rt   ru   )Ú	alphacumsr{   Úetary   ri   Úalphas_prevÚsigmass   &&&&   r   Úmake_ddim_sampling_parametersr„   Š   s¶   € àÕ&€FÜ—*’*˜i¨�l˜^¨iÀsÈÐ8KÕ.L×.SÑ.SÓ.UÕUÓV€Kð ”2—7’7˜A �O°°Fµ
Õ;¸qÀ6ÕCWÕ?WÕXÓYÕY€FßÜ�ŠÐ>¸v¸hÀkÐR]ÐQ^Ð_Ô`Ü�ŠÐ=¸c¸Uð CQØQWÐPXðZô 	[à˜;Ð&Ð&r   c           
     óð   € . p\        V 4       FP  pW@,          pV^,           V ,          pVP                  \        ^V! V4      V! V4      ,          ,
          V4      4       KR  	  \        P                  ! V4      # )aÿ  
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].
:param num_diffusion_timesteps: the number of betas to produce.
:param alpha_bar: a lambda that takes an argument t from 0 to 1 and
                  produces the cumulative product of (1-beta) up to that
                  part of the diffusion process.
:param max_beta: the maximum beta to use; use values lower than 1 to
                 prevent singularities.
)rq   ÚappendrM   r^   Úarray)Únum_diffusion_timestepsÚ	alpha_barÚmax_betarg   ÚiÚt1Út2s   &&&    r   ra   ra   ˜   se   € ð €EÜÐ*Ö+ˆØÕ(ˆØ�!�eÐ.Õ.ˆØ�‰”S˜™Y r›]©Y°r«]Õ:Õ:¸HÓEÖFñ ,ô �8Š8�E‹?Ðr   c                 ó–   € VP                   vr4V P                  RV4      pVP                  ! V.R\        V4      ^,
          ,          O5!  # )rK   rZ   ©rK   )ÚshapeÚgatherÚreshapeÚlen)ÚarU   Úx_shapeÚbÚ_Úouts   &&&   r   Úextract_into_tensorr™   «   s>   € Ø�G‰G€E€AØ
�(‰(�2�q‹/€CØ�;Š;�qÐ7˜D¤C¨£L°1Õ$4Õ5Ó7Ð7r   c                ó˜   € V'       d>   \        V4      \        V4      ,           p\        P                  ! V \        V4      .VO5!  # V ! V!  # )a›  
Evaluate a function without caching intermediate activations, allowing for
reduced memory at the expense of extra compute in the backward pass.
:param func: the function to evaluate.
:param inputs: the argument sequence to pass to `func`.
:param params: a sequence of parameters `func` depends on but does not
               explicitly take as arguments.
:param flag: if False, disable gradient checkpointing.
)ÚtupleÚCheckpointFunctionÚapplyr“   )ÚfuncÚinputsÚparamsÚflagÚargss   &&&& r   Ú
checkpointr£   ±   s?   € ÷ Ü�V‹}œu V›}Õ,ˆÜ!×'Ò'¨¬c°&«kÐA¸DÓAÐAá�V‰}Ðr   c                   ó@   a € ] tR t^Ât o ]R 4       t]R 4       tRtV tR# )rœ   c                ó�  € Wn         \        VR V 4      V n        \        W2R  4      V n        R\        P
                  ! 4       R\        P                  ! 4       R\        P                  ! 4       /V n        \        P                  ! 4       ;_uu_ 4        V P                   ! V P                  !  pR R R 4       V#   + '       g   i     X# ; i)NÚenabledr8   Úcache_enabled)
Úrun_functionrp   Úinput_tensorsÚinput_paramsr   Úis_autocast_enabledÚget_autocast_gpu_dtypeÚis_autocast_cache_enabledÚgpu_autocast_kwargsÚno_grad)Úctxr¨   Úlengthr¢   Úoutput_tensorss   &&&* r   r<   ÚCheckpointFunction.forwardÃ   s�   € à'ÔÜ   g v Ó/ˆÔÜ  W Ó.ˆÔØ#,¬e×.GÒ.GÓ.IØ#*¬E×,HÒ,HÓ,JØ#2´E×4SÒ4SÓ4Uð#WˆÔô �]Š]�_�_Ø ×-Ò-¨s×/@Ñ/@ÑAˆN÷ àÐ÷ Ž_àÐús   ÂB4Â4C	c           
     ó¼  € V P                    Uu. uF!  q"P                  4       P                  R 4      NK#  	  upV n         \        P                  ! 4       ;_uu_ 4        \        P
                  P                  P                  ! R/ V P                  B ;_uu_ 4        V P                    Uu. uF  q"P                  V4      NK  	  ppV P                  ! V!  pRRR4       RRR4       \        P                  P                  XV P                   V P                  ,           VR R7      pV = V =?RV,           # u upi u upi   + '       g   i     Ln; i  + '       g   i     Ly; i)TN)Úallow_unused© )NN)r©   ÚdetachÚrequires_grad_r   Úenable_gradÚcudaÚampÚautocastr®   Úview_asr¨   ÚautogradÚgradrª   )r°   Úoutput_gradsr;   Úshallow_copiesr²   Úinput_gradss   &*    r   ÚbackwardÚCheckpointFunction.backwardÏ   s  € àFI×FWÒFWÓXÑFWÀŸX™X›Z×6Ñ6°tÖ<ÑFWÑXˆÔÜ×Ò× Õ Ü—
‘
—‘×'Ò'ÑB¨#×*AÑ*A×BÓBð 58×4EÒ4EÓFÑ4E¨qŸi™i¨žlÑ4EˆNÐFØ ×-Ò-¨~Ñ>ˆN÷ C÷ !ô —n‘n×)Ñ)ØØ×Ñ × 0Ñ 0Õ0ØØð	 *ó 
ˆð ÐØÐØØ˜kÕ)Ð)ùò# Yùò G÷	 C×Bú÷ !× ús;   �'D.Á=EÂD8	Â'D3Ã D8	ÃEÄ3D8	Ä8EÅEÅE	r¶   N)	r?   r@   rA   rB   Ústaticmethodr<   rÃ   rC   rD   )r   s   @r   rœ   rœ   Â   s+   ø‡ € Øñ	ó ð	ð ñ*ó ö*r   rœ   c           
     ór  € V'       Eg    V^,          p\         P                  ! \        P                  ! V4      ) \         P                  ! ^ V\         P
                  V P                  R7      ,          V,          4      pV R,          P                  4       VR,          ,          p\         P                  ! \         P                  ! V4      \         P                  ! V4      .RR7      pV^,          '       d6   \         P                  ! V\         P                  ! VR,          4      .RR7      pV# \        V RVR7      pV# )	a<  
Create sinusoidal timestep embeddings.
:param timesteps: a 1-D Tensor of N indices, one per batch element.
                  These may be fractional.
:param dim: the dimension of the output.
:param max_period: controls the minimum frequency of the embeddings.
:return: an [N x dim] Tensor of positional embeddings.
)ÚstartÚendr8   r-   N©Údimzb -> b d)Úd)ºNNNNrZ   )rÌ   :NrK   N)r   ÚexprR   Úlogr]   Úfloat32r-   r   ÚcatrS   ÚsinÚ
zeros_liker   )rh   rÊ   Ú
max_periodÚrepeat_onlyÚhalfÚfreqsr¢   Ú	embeddings   &&&&    r   Útimestep_embeddingrØ   å   sä   € ÷ ˆ;Ø�a�xˆÜ—	’	Ü�XŠX�jÓ!Ð!¤E§L¢L°q¸dÌ%Ï-É-Ð`i×`pÑ`pÔ$qÕqÐtxÕxó
ˆð ˜Õ!×'Ñ'Ó)¨E°$­KÕ7ˆÜ—I’IœuŸyšy¨›´·	²	¸$³Ð@ÀbÔIˆ	Ø��7Œ7ÜŸ	š	 9¬e×.>Ò.>¸yÈÕ?OÓ.PÐ"QÐWYÔZˆIð Ðô ˜9 j°CÔ8ˆ	ØÐr   c                óp   € V P                  4        F!  pVP                  4       P                  4        K#  	  V # )z4
Zero out the parameters of a module and return it.
)Ú
parametersr·   Úzero_)ÚmoduleÚps   & r   Úzero_modulerÞ   ü   s.   € ð ×ÑÖ ˆØ	�‰‹
×ÑÖñ !à€Mr   c                ór   € V P                  4        F"  pVP                  4       P                  V4       K$  	  V # )z1
Scale the parameters of a module and return it.
)rÚ   r·   Úmul_)rÜ   ÚscalerÝ   s   && r   Úscale_modulerâ     s.   € ð ×ÑÖ ˆØ	�‰‹
�‰˜Öñ !à€Mr   c           
     ór   € V P                  \        \        ^\        V P                  4      4      4      R7      # )z.
Take the mean over all non-batch dimensions.
rÉ   )Úmeanrp   rq   r“   r�   )Útensors   &r   Ú	mean_flatræ     s*   € ð �;‰;œ4¤ a¬¨V¯\©\Ó):Ó ;Ó<ˆ;Ó=Ð=r   c                óÌ   € V ^8X  d   \         P                  ! V/ VB # V ^8X  d   \         P                  ! V/ VB # V ^8X  d   \         P                  ! V/ VB # \	        RV  24      h)z0
Create a 1D, 2D, or 3D average pooling module.
zunsupported dimensions: )r!   Ú	AvgPool1dÚ	AvgPool2dÚ	AvgPool3dr#   )Údimsr¢   Úkwargss   &*,r   Úavg_pool_ndrí     sh   € ð ˆq„yÜ�|Š|˜TÐ, VÑ,Ð,Ø	�ŒÜ�|Š|˜TÐ, VÑ,Ð,Ø	�ŒÜ�|Š|˜TÐ, VÑ,Ð,Ü
Ð/°¨vÐ6Ó
7Ð7r   c                   ó8   a a€ ] tR tRt oV 3R ltR tRtVtV ;t# )ÚHybridConditioneri"  c                ód   <€ \         SV `  4        \        V4      V n        \        V4      V n        R # r>   )r   r   r   Úconcat_conditionerÚcrossattn_conditioner)r$   Úc_concat_configÚc_crossattn_configr%   s   &&&€r   r   ÚHybridConditioner.__init__$  s)   ø€ Ü‰ÑÔÜ"9¸/Ó"JˆÔÜ%<Ð=OÓ%PˆÖ"r   c                óV   € V P                  V4      pV P                  V4      pR V.RV./# )Úc_concatÚc_crossattn©rñ   rò   )r$   r÷   rø   s   &&&r   r<   ÚHybridConditioner.forward)  s5   € Ø×*Ñ*¨8Ó4ˆØ×0Ñ0°Ó=ˆØ˜X˜J¨¸°}ÐEÐEr   rù   )	r?   r@   rA   rB   r   r<   rC   rD   rE   rF   s   @@r   rï   rï   "  s   ù‡ € õQ÷
Fò Fr   rï   c                 óN   a a€ VV 3R  lpVV 3R lpV'       d   V! 4       # V! 4       # )c                  óª   <€ \         P                  ! ^.SR,          O5S R7      P                  ! S^ ,          .R\        S4      ^,
          ,          O5!  # )rK   rJ   r,   r�   )r   Úrandnr   r“   ©r-   r�   s   €€r   rV   Únoise_like.<locals>.<lambda>0  sB   ø€ œ5Ÿ;š;¨ ¨E°"­I¡¸vÔF×MÒMÈeÐTUÍhÐsÐZ^ÔbeÐfkÓblÐopÕbpÕZqÔsr   c                  ó4   <€ \         P                  ! SS R 7      # )r,   )r   rý   rþ   s   €€r   rV   rÿ   1  s   ø€ ”E—K’K ¨fÕ5r   r¶   )r�   r-   r   Úrepeat_noiseÚnoises   ff&  r   Ú
noise_liker  /  s   ù€ Ýs€LÝ5€Eß#‰<‹>Ð0©«Ð0r   )g-Cëâ6?g{®Gáz”?rQ   )T)rL   )i'  F)F)rR   rt   r   Útorch.nnr!   Únumpyr^   Úeinopsr   r   Úcomfy.ldm.utilr   ÚModuler   rj   r}   r„   ra   r™   r£   r¾   ÚFunctionrœ   rØ   rÞ   râ   ræ   rí   rï   r  r¶   r   r   Ú<module>r
     s•   ðó Û Û Ý Û ß $å 2ôB�2—9‘9ô BôJô@ô"'ôò&8òô" *˜Ÿ™×0Ñ0ô  *ôFò.òò>ò
8ô
F˜Ÿ	™	ô 
Fö1r   