]iH*SrSSKJr SSKrSSKrSSKrSSKJr SSK J r SSK r SSK Jr SSKJr SS KJr SS KJrJr \R,\\\4r"S S \R05r"S S5rSrSS\R,\\4S\R,\R8\4S\S\S\S\S\S\S\R@\S\R84Sjjr!g)z Code to apply a model to a mix. It will handle chunking with overlaps and inteprolation between chunks, as well as the "shift trick". )ThreadPoolExecutorN)nn) functional)Demucs)HDemucs)HTDemucs) center_trimDummyPoolExecutorc^\rSrSrS S\R \S\R\R \R \S\R\4U4Sjjjr \ S\4Sj5r Sr S r U=r$) BagOfModelsmodelsweightssegmentc>[TU]5 [U5S:deUSnUHbnURUR:XdeURUR:XdeUR UR :XdeUcM\X5lMd UR UlURUlURUl[R"U5Ul Uc-UVs/sHodRVs/sHnSPM snPM! nnOG[U5[U5:XdeUH'n[U5[UR5:XaM'e X l gs snfs snf)aJ Represents a bag of models with specific weights. You should call `apply_model` rather than calling directly the forward here for optimal performance. Args: models (list[nn.Module]): list of Demucs/HDemucs models. weights (list[list[float]]): list of weights. If None, assumed to be all ones, otherwise it should be a list of N list (N number of models), each containing S floats (S number of sources). segment (None or float): overrides the `segment` attribute of each model (this is performed inplace, be careful is you reuse the models passed). rN?) super__init__lensources samplerateaudio_channelsrr ModuleListrr) selfrrrfirstother_weight __class__s I/mnt/rpi/tmp/demucs-venv-sys/lib/python3.13/site-packages/demucs/apply.pyrBagOfModels.__init__s4 6{Qq E==EMM1 11##u'7'77 77''5+?+?? ??" ' $22**}} mmF+ ?%++e,%%N#U##AAr,r c4\rSrSrSSjr\S5rSrSrg) TensorChunkONc&URSnUS:deX$:deUcXB- nO[XB- U5n[U[5(a%URUlX!R -UlO XlX lX0lURUlg)Nr)shaper(r'r>tensoroffsetlengthdevice)rrCrDrE total_lengths r!rTensorChunk.__init__Ps||B' {{$$$ >!*F.7F fk * * --DK ==0DK K K mm r,cb[URR5nURUS'U$)NrA)listrCrBrE)rrBs r!rBTensorChunk.shapecs(T[[&&'KKb  r,cHXR- nURRSnUS:deURUS-- nXA-n[ SU5n[ X55nXd- nXW- n [ R"URSXg24X45n U RSU:XdeU $)NrAr.)rErCrBrDmaxr(Fpad) r target_lengthdeltarGstartend correct_start correct_endpad_left pad_rightouts r!paddedTensorChunk.paddedis +{{((, zz eqj(#Au ,,  (% eeDKK]%> >?(AVWyy} --- r,)rFrErDrC)rN) r2r3r4r5rr:rBrZr;r,r!r>r>Os $& r,r>c[U[5(aU$[U[R5(de[U5$)N)r'r>thTensor)tensor_or_chunks r! tensor_chunkra|s7/;///2995555?++r,r*mixshiftssplitoverlaptransition_powerprogress num_workersrr#c ~ Uc URnO[R"U5nU c,US:aURS:Xa [U5n O [ 5n UUUUUUU U S.n [ U[ 5(Ga9Sn S/[UR5-n [URUR5Hup[[UR555RnURU5 [!X40U D6nURU5 [#U5H*unnUSS2USS2SS24==U-ss'U U==U- ss'M, U U- n AM [ U [R$5(de['U R(S5HnU SS2USS2SS24==U U-ss'M U $URU5 UR+5 US:dS5eUR(unnnU(aSU S'[-S UR.-5n[1U5n[ U[25(deUR5US U--5nSn['U5HIn[6R8"SU5n[3UUUU-U- 5n[!UU40U D6nUUS UU- S24- nMK UU-n[ U[R$5(deU$U(GaoS U S '[R:"U[UR5UUURS9n[R:"UURS9nU c UR<n U bU S:de[-UR.U -5n[-SU- U-5n['SUU5n [?[AUUR.- S55n![RB"[RD"SUS -S-US9[RD"UUS -- SSUS9/5n"[U"5U:XdeU"U"RG5- U-n"/n#U HAn[3UUU5n$U RH"[ UU$40U D6n%U#RKU%U45 UU- nMC U(a[LRL"U#U!SSS9n#U#Hun%nU%RO5n&U&R(Sn'US UUU-24==U"SU'U&-RUR5- ss'UUUU-===U"SU'RUR5- sss&M URQ5S:deUU-n[ U[R$5(deU$[ U[R5(aU b[-XR.-5n(O%[UUS5(aURWU5n(OUn([1U5n[ U[25(deUR5U(5RU5n[RX"5 U"U5nSSS5 [ W[R$5(de[[UU5$!,(df  N;=f)aa Apply model to a given mixture. Args: shifts (int): if > 0, will shift in time `mix` by a random amount between 0 and 0.5 sec and apply the oppositve shift to the output. This is repeated `shifts` time and all predictions are averaged. This effectively makes the model time equivariant and improves SDR by up to 0.2 points. split (bool): if True, the input will be broken down in 8 seconds extracts and predictions will be performed individually on each and concatenated. Useful for model with large memory footprint like Tasnet. progress (bool): if True, show a progress bar (requires split=True) device (torch.device, str, or None): if provided, device on which to execute the computation, otherwise `mix.device` is assumed. When `device` is different from `mix.device`, only local computations will be on `device`, while the entire tracks will be stored on `mix.device`. num_workers (int): if non zero, device is 'cpu', how many threads to use in parallel. segment (float or None): override the model segment parameter. Nrcpu)rcrdrerfrgrFpoolrgrz-transition_power < 1 leads to weird behavior.rcg?rM.Frd)rFz.2frAxseconds) unit_scalencolsunit valid_length).rFr^typerr r'r rrziprrnextiter parametersto apply_model enumerater_rangerBevalintrrar>rZrandomrandintzerosrr&formatcatarangerNsubmitappendtqdmresultr(r hasattrrqno_gradr ))r*rbrcrdrerfrgrFrhrrkkwargs estimatestotals sub_model model_weightsoriginal_model_devicerYk inst_weightbatchchannelsrE max_shift padded_mixrrDshifted shifted_out sum_weightsegment_lengthstrideoffsetsscalerfutureschunkfuture chunk_out chunk_lengthrqs) r!rxrxs6~6" | ?v{{e3%k2D$&D, $F%%%13 EMM**(+ELL%--(H $I$(i.B.B.D)E$F$M$M ! LL i77C LL. /"+M":;Aq!QJ;.q [( #;  I)I)RYY////yq)*A aAqj !VAY . !+ HHV JJL q Q"QQ !iiE8V xe.../ 3#{++++ZZY 67 vA^^Ay1F!*ffy6H66QRG%eW??K ;sI$6$778 8C  v #ryy)))) whhuc%--0(F3::VXXfSZZ8 ?mmG"w|33!%"2"2W"<=a'k^3466*fVe&6&66>?1n&9A&=fM>Na4G#GBW]^`a6{n,,,6::<'*::FV^ B 4 vf~5 6&,:O:R:RSVS]S]:^ ^ 6 & ~~!### z#ryy))))  eX & &7+>w)9)99:L UN + + --f5L!L3#{++++ZZ -008 ZZ\ #C#ryy))))3''\s 0 X.. X<) rTg?rFNrNN)"__doc__concurrent.futuresrr}typingr6torchr^rtorch.nnrrOrdemucsrhdemucsrhtdemucsr utilsr r Unionr8Moduler r>rar_r|boolr&r9rxr\r,r!rs 2 $ 1 (*+1A"))1Ah**Z,04AC/3DH M(rxx U 23M(XXbii45M(M((,M(M(:?M( M( ! M( 02{{5/A M( !ii M(r,