Vh 0ddlZddlZddlmZddlmZmZmZmZm Z m Z ddl Z ddl Z ddl Z ddlZdedefdZ dPdeded eed eefd Zd efd ZGddeZdQde jdedede jfdZde jde jfdZde jde jfdZGdde jjjZdej dej!deeeej!ffdZ"GddeZ#Gdd eZ$Gd!d"eZ%Gd#d$eZ&Gd%d&eZ'e(d'krfej d()Z)e)*d*ed+gd,d-.e)*d/ed01e)*d2d3d45e)*d6ed78e)*d9e+d:8e)*d;ede)*d?ed@dA>e)*dBe+dC8e),\Z-Z.e j/e-j0e"e)e-\Z1Z2Z-e3dDe j4dZ5eje6e7e1D]Z8e1e8\Z9Z:e5e9j;dEe1j<z z Z5e-j=rVe j=dFee8zdGze9j>e1j<e j=dFee8zdHze:j>e1j<e3dIe5dJz e3dKe7e1e3dLe7e2e-j?e1_@e jjAe1e-jBdMdNOZCejeCD]\Z9Z:e3e9j;dSdS)RN)Path)OptionalUnionTupleListAnyCallablepathreturnctjdkrtdi}tjt |}|j|d<|j|d<|j|d<|d|dz |d<|S)zLoad audio metadata this is a backend_independent wrapper around torchaudio.info Args: path: Path of filename Returns: Dict: Metadata with `samplerate`, `samples` and `duration` in seconds soxz#Deprecated backend is not supported sampleratesampleschannelsduration) torchaudioget_audio_backend RuntimeErrorinfostr sample_rate num_frames num_channels)r rsis T/var/www/html/movieo_spanner_bot/venv/lib/python3.11/site-packages/openunmix/data.py load_infor s#%%..@AAA D T # #BDmDODIl);;D Kstartdurrc|tj|\}}||fS|t|}t||dz}t||dz}tj|||\}}||fS)a`Load audio file Args: path: Path of audio file start: start position in seconds, defaults on the beginning. dur: end position in seconds, defaults to `None` (full file). info: metadata object as called from `load_info`. Returns: Tensor: torch tensor waveform of shape `(num_channels, num_samples)` Nr)r frame_offset)rloadrint)r rr rsigraterr"s r load_audior'%s$ {OD)) TDy <T??DtL1122 54 #5566 ODZl[[[ TDyrlist_of_function_namesc>|rtd|DSdS)Nc<g|]}td|zS _augment_globals.0augs r z aug_from_str..Gs&WWW +"34WWWrc|SNaudios rzaug_from_str..IsUr)Compose)r(s r aug_from_strr:Es1#WW@VWWWXXX"""rc<eZdZdZdZdejdejfdZdS)r9zqComposes several augmentation transforms. Args: augmentations: list of augmentations to compose. c||_dSr4 transforms)selfr>s r__init__zCompose.__init__Rs $rr7r c0|jD] }||}|Sr4r=)r?r7ts r__call__zCompose.__call__Us'  AAeHHEE rN)__name__ __module__ __qualname____doc__r@torchTensorrCr5rrr9r9LsS %%%elu|rr9??r7lowhighcF|tjd||z zz}||zS)z.Applies a random gain between `low` and `high`)rHrand)r7rLrMgs r _augment_gainrR[s' ejmmtcz**A 19rc|jddkr@tjddkrtj|dgS|S)z;Swap channels of stereo signals with a probability of p=0.5rg?g?)shaperHtensoruniform_flipr6s r_augment_channelswaprYasL {1~u|C0099;;cAAz%!%%% rc|jddkr |dddf}|jddkrtj|dd}|S)NrrT.rOdim)rUrHrepeat_interleaver6s r_augment_force_stereor^isQ {1~bqb#g {1~'qa888 Lrc eZdZdZ ddeeefdedeedee ddf dZ d e de fd Z de fd Zdefd Zdefd ZdS) UnmixDatasetNrootr seq_durationsource_augmentationsr cttj|_||_||_||_dSr4)rargsrb expanduserrrcrd)r?rbrrcrds rr@zUnmixDataset.__init__xs=OO..00 &($8!!!rindexctr4NotImplementedError)r?rhs r __getitem__zUnmixDataset.__getitem__!!rctr4rjr?s r__len__zUnmixDataset.__len__rmrcdjjz}dg}|z }|gfd|Dz}d|S)NzDataset zNumber of datapoints: {}c*g|]}djz|zS) ) _repr_indent)r0liner?s rr2z)UnmixDataset.__repr__..s&JJJT# 11D8JJJr ) __class__rDformatrp extra_repr splitlinesjoin)r?headbodyliness` r__repr__zUnmixDataset.__repr__sDN33*11$,,..AAB !!,,...JJJJTJJJJyyrcdS)Nr5ros rryzUnmixDataset.extra_reprsrr)NN)rDrErFrtrrrfloatrr r@r$rrlrprryr5rrr`r`usL )-37 9 9D#I 9 9uo 9 'x0 9  9 9 9 9""""""""""" #    Crr`parserrfc  |jdkr|dt|dt|}t |jj|_t |j|j |j |jd}t |jj|_td:ddd|}td:d d i|}nd|jd krB|d td |dt|dtd|dtd|dtd|dtd |}|j |_t |j|j|j |jd}t!|j}t%d:d|d|j|j d|}t%d:d d|j |jd|}n|jdkr|dt|dtd |ddd d!"|dtd |}t |jj|_t |j|j|jd#}t!|j}t/d:d||jd|j d$|}t/d:d d%d&|}n|jd'kr,|dtd|dt|dtd |d(dd d)"|d*dd d+"|}t |jj|_t |j|j|j|jd,}t5d-|jD}t7d:d||jd|j d.|}t7d:d d%d&|}n|d/dd d0"|d1td2|dtd3d4gd 5|}|j|jd|j|jd%u|jd6}t!|j}t?d:d|j |j |dd7|}t?d:d d8d%d9|}|||fS);zsLoads the specified dataset from commandline arguments Returns: train_dataset, validation_dataset alignedz --input-file)typez --output-file)rbrc input_file output_filetrainT)split random_chunksrvalid sourcefolderz--interferer-dirs+)rnargsz --target-dirz--ext.wavrdefaultz--nb-train-samplesz--nb-valid-samplesdz--source-augmentations)rbinterferer_dirs target_dirext)rrdr nb_samplesrc)rrrcrtrackfolder_fixz --target-filez--interferer-filesz--random-track-mix store_trueFz&Apply random track mixing augmentation)actionrhelp)rbinterferer_files target_file)rrdrandom_track_mixrrcN)rrctrackfolder_varz--random-interferer-mixz+Apply random interferer mixing augmentationz--silence-missingzsilence missing targets)rbrrsilence_missing_targetsc<g|]}td|zSr+r-r/s rr2z!load_datasets..s&'j'j'j +2C(D'j'j'jr)rrdrandom_interferer_mixrrcz--is-wavzloads wav instead of STEMSz--samples-per-track@gain channelswap)rrr)rbis_wavsubsetstargetdownloadseed)rsamples_per_trackrcrdrrO)rrrcr5)!dataset add_argumentr parse_argsrrstemrrbseq_durrAlignedDatasetr$rrrr:rdSourceFolderDatasetnb_train_samplesnb_valid_samplesrrFixedSourcesTrackFolderDatasetrsilence_missingr9!VariableSourcesTrackFolderDatasetrrr MUSDBDatasetr)rrfdataset_kwargs train_dataset valid_datasetrds r load_datasetsrs5 |y  N555O#666  ""4+,,1 OO L/+    4+,,1 &[WD[[N[[ &GGWGGG  ' '/cEEEN555G#v>>>0sDIII0sCHHH43cJJJ  ""o OO#3/8    ,D,EFF+ !5,      , ,       * * *O#6660s#FFF 9    43cJJJ  ""4+,,1 OO $ 5+   ,D,EFF6 !5!2      7jWSWjj[ijj * * *G#v>>>O#66643cJJJ %>     *      ""4+,,1 OO+8'+';    ''j'jPTPi'j'j'jkk9 !5"&"<      :mVZmm^lmm   -    1RHHH43Q^H_gjkkk  ""Ikk T)I    ,D,EFF$ "4!5!      %m7aVZmm^lmm - --rczeZdZ dded ed ed ed eed ededeededdfdZ dZ dZ dZ dS)rr mixture.wav vocals.wavNF@*rbrrrrcrrrdrr c ^t||_||_||_||_||_||_||_t| |_ |j std| |_ tj |j dS)aA dataset of that assumes multiple track folders where each track includes and input and an output file which directly corresponds to the the input and the output of the model. This dataset is the most basic of all datasets provided here, due to the least amount of preprocessing, it is also the fastest option, however, it lacks any kind of source augmentations or custum mixing. Typical use cases: * Source Separation (Mixture -> Target) * Denoising (Noisy -> Clean) * Bandwidth Extension (Low Bandwidth -> High Bandwidth) Example ======= data/train/01/mixture.wav --> input data/train/01/vocals.wav ---> output z)Dataset is empty, please check parametersN)rrgrbrrrcrrrlist _get_paths tuple_pathsrrrandom) r?rbrrrrcrrrdrs rr@zAlignedDataset.__init__?s@JJ))++  &(*$& 1 122 LJKK K  DIrcT|j|\}}|jrXt|}t|}t|d|d}t jd||jz }nd}t|||j\}} t|||j\} } || fS)Nrrrr )rrrminruniformrcr') r?rh input_path output_path input_info output_inforrX_audio_Y_audios rrlzAlignedDataset.__getitem__ms"&"25"9 K   ":..J#K00K:j1;z3JKKHN1h1B&BCCEEE %T=NOOO  5d>OPPP rc*t|jSr4)lenrros rrpzAlignedDataset.__len__}s4#$$$rc#VKt|j|j}tj|D]}|rt ||j}t ||j }|r|r}|j dt|d}t|d}t|d|d}||j kr|d|dfV|d|dfVdS)Loads input and output tracksNrr) rrbrtqdmiterdiris_dirrglobrrrcrr)r?p track_pathrrrr min_durations rrzAlignedDataset._get_pathss$ DJ ' ')AIIKK00 < s''(,#$37,,,, ,  , uo ,,,'x0,, ,,,,\   %%%<<<<rrrbrrrrrrcrrrdrr c lt||_||_| |_||_||_||_| |_||_ ||_ |j |j gz|_ | |_ ||_| |_t!j|jdS)aA dataset that assumes folders of sources, instead of track folders. This is a common format for speech and environmental sound datasets such das DCASE. For each source a variable number of tracks/sounds is available, therefore the dataset is unaligned by design. By default, for each sample, sources from random track are drawn to assemble the mixture. Example ======= train/vocals/track11.wav ----------------- train/drums/track202.wav (interferer1) ---+--> input train/bass/track007a.wav (interferer2) --/ train/vocals/track11.wav ---------------------> output N)rrgrbrrrcrrrdrrsource_folders get_tracks source_tracksrrr) r?rbrrrrrrcrrrdrs rr@zSourceFolderDataset.__init__s@JJ))++  &(*$8!$."2do5FF!__..$  DIrc8g}|jD]}|jdkrtj|tj|j|}t |d}|jrtjd||j z }nt|dz|j dzz d}t|||j \}}| |}| |tj|} | d} | d} | | fS)NrrrrTr)rrrrchoicerrrrrcmaxr'rdappendrHstacksum) r?rh audio_sourcessource source_pathrrr7rstemsxys rrlzSourceFolderDataset.__getitem__s! ) ( (FzW$$ E"""!-(:6(BCCK --j9H! Gq(T5F*FGGHMD,=,BBAFF!+U@QRRRHE1--e44E   ' ' ' ' M** IIaLL "I!t rc|jSr4)rros rrpzSourceFolderDataset.__len__s rct|j|j}i}tj|jD]}g}||z }t |d|jzD]T}|j6t|}|d|jkr| |?| |U|||<|S)r*Nr) rrbrrrsortedrrrcrr)r?rr source_foldertracksrsource_track_pathrs rrzSourceFolderDataset.get_trackss DJ ' ' !Yt':;; 2 2MFm+K%+K,<,rrrrrbrrrrcrrrdrrr c t||_||_| |_||_||_||_||_||_ ||_ |j |j gz|_ | |_ tj |j t||_t#|jst%ddS)aA dataset that assumes audio sources to be stored in track folder where each track has a fixed number of sources. For each track the users specifies the target file-name (`target_file`) and a list of interferences files (`interferer_files`). A linear mix is performed on the fly by summing the target and the inferers up. Due to the fact that all tracks comprise the exact same set of sources, the random track mixing augmentation technique can be used, where sources from different tracks are mixed together. Setting `random_track_mix=True` results in an unaligned dataset. When random track mixing is enabled, we define an epoch as when the the target source from all tracks has been seen and only once with whatever interfering sources has randomly been drawn. This dataset is recommended to be used for small/medium size for example like the MUSDB18 or other custom source separation datasets. Example ======= train/1/vocals.wav --------------- train/1/drums.wav (interferer1) ---+--> input train/1/bass.wav -(interferer2) --/ train/1/vocals.wav -------------------> output zNo tracks foundN)rrgrbrrrcrrrdrr source_filesrrrrrrr) r?rbrrrrcrrrdrrs rr@z'FixedSourcesTrackFolderDataset.__init__sTJJ))++  &( 0*$8!& 0 1T5E4FF  DI4??,,-- 4; 2011 1 2 2rcL|j|d}|j|d}|jrtjd||jz }nd}g}t ||jz ||j\}}||}|||j D]}|j r}tj tt|j} |j| d}|jr0|j| d}tjd||jz }t ||z ||j\} }|| } || tj|} | d} | d} | | fS)Nr rrr)rrrrrcr'rrdrrrrrangerrHrr)r?rhrrrr target_audiorr random_idxr7rrrs rrlz*FixedSourcesTrackFolderDataset.__getitem__4s['/ {5).9   N1lT5F&FGGEEE $Z$2B%B%UYUfggg a00>> \***+ ( (F$ P#]5T[1A1A+B+BCC ![4V< %P#';z#:>#JL"N1lT=N.NOOE!*v"5UHYZZZHE1--e44E   ' ' ' ' M** IIaLL !H!t rc*t|jSr4rrros rrpz&FixedSourcesTrackFolderDataset.__len__Y4;rc#Kt|j|j}tj|D]rfd|jD}td|DstdS|j Nttt|}td|D}||j kr|dVddVdS)rcg|]}|z Sr5r5)r0srs rr2z=FixedSourcesTrackFolderDataset.get_tracks..asJJJ1 QJJJrc3>K|]}|VdSr4)exists)r0sps r z.bs*>>2299;;>>>>>>rzExclude track Nc3&K|] }|dV dSrNr5r0is rrz.is&&D&Dq}&D&D&D&D&D&Drr r)rrbrrrrrallprintrcrmaprr)r?r source_pathsinfosrrs @rrz)FixedSourcesTrackFolderDataset.get_tracks\s! DJ ' ')AIIKK00 G GJ  "" GJJJJ8IJJJ >>>>>>>*J777$0 Y !=!=>>E#&&D&De&D&D&D#D#DL#d&777(2LQQRRR$.EEFFF G Gr)rDrErFrrrrrr r$r@rlrprr5rrrrs''1;&?(,#!&3F3F$:2:2:2:2 :2 s) :2 uo :2:2:2'x0:2:2:2 :2:2:2:2x###J   GGGGGrrceZdZdddddddddf ded ed ed ed eed edededeededdfdZdZ dZ dZ dS)rrrrNFrc|Sr4r5r6s rr8z*VariableSourcesTrackFolderDataset.{rrrbrrrrcrrrrdrr c t||_||_||_||_||_||_| |_||_ ||_ | |_ t| |_dS)aA dataset that assumes audio sources to be stored in track folder where each track has a _variable_ number of sources. The users specifies the target file-name (`target_file`) and the extension of sources to used for mixing. A linear mix is performed on the fly by summing all sources in a track folder. Since the number of sources differ per track, while target is fixed, a random track mix augmentation cannot be used. Instead, a random track can be used to load the interfering sources. Also make sure, that you do not provide the mixture file among the sources! Example ======= train/1/vocals.wav --> input target train/1/drums.wav --> input target | train/1/bass.wav --> input target --+--> input train/1/accordion.wav --> input target | train/1/marimba.wav --> input target / train/1/vocals.wav -----------------------> output N)rrgrbrrrcrrrdrrrrrr) r?rbrrrrcrrrrdrs rr@z*VariableSourcesTrackFolderDataset.__init__qsNJJ))++  &(*%:"$8!&'>$4??,,-- rc*|j|d}|jr1|j|d}tjd||jz }nd}|jrtjtt|j}|j|d}|jr1|j|d}tjd||jz }nd}n|}|}tt| d|j z} d} | D]|} | ||j z kr t| ||j\} } n7#t$r*|dkr|dz n|dz}||cYcSwxYw| || z } }t%||j z r=t||j z ||j\}} ||}| |z } nt)j| j}| |fS)Nr rrrrrO)rrrrrcrrrrrrrrrr'rrlrdrr rHzerosrU)r?rhtarget_track_pathtarget_min_duration target_startrintfr_track_pathintfr_min_duration intfr_startsourcesrrr7rrs rrlz-VariableSourcesTrackFolderDataset.__getitem__sW K.v6   "&+e"4^"D !>!-@4CT-TUULLL  % 'uS-=-='>'>??J#{:6v> ! %)[%<^%L"$nQ0BTEV0VWW  0 &K.33C$(NCCDDEE " 2 2K.1AAAA /%k$J[\\\qq / / /%*QYY EAI''....... / **511 1AA !D$44 5 5 < < > > )!D$44"%DAq ))!,,A FAA EK((A!t s'E/E65E6c*t|jSr4rros rrpz)VariableSourcesTrackFolderDataset.__len__rrc #`Kt|j|j}tj|D]}|rt||js|jrtt| d|j z}|std||jNttt |}t#d|D}||jkr||dV|ddVdS)Nrz empty track: c3&K|] }|dV dSrr5rs rrz?VariableSourcesTrackFolderDataset.get_tracks..s&*H*HQ1Z=*H*H*H*H*H*Hrr)rrbrrrrrr rrrrrrrcrrr)r?rrr$rrs rrz,VariableSourcesTrackFolderDataset.get_trackss@ DJ ' ')AIIKK00 K KJ  "" K D$455<<>>K$B^K$T*//#.*I*I%J%JKKG"!oz::: (4 $SG%<%< = ='**H*H%*H*H*H'H'H '$*;;;,6 #U#UVVV(2DIIJJJ# K Kr) rDrErFrrrrr r@rlrprr5rrrrps'(,#&+$3F3F(-1.1.1.1. 1.  1. uo 1.1. $1.1.'x01."&1. 1.1.1.1.f555n   KKKKKrrceZdZdddddddddddf d ed ed ed ed ededeededeedededdfdZ dZ dZ dS)rrNFrg@rc|Sr4r5r6s rr8zMUSDBDataset.rrrrrbrrrrrcrrdrrr c ddl}| |_tj| ||_||_||_||_||_||_| |_ | |_ |j | |||||d| |_ d|_ dS)aMUSDB18 torch.data.Dataset that samples from the MUSDB tracks using track and excerpts with replacement. Parameters ---------- target : str target name of the source to be separated, defaults to ``vocals``. root : str root path of MUSDB download : boolean automatically download 7s preview version of MUSDB is_wav : boolean specify if the WAV version (instead of the MP4 STEMS) are used subsets : list-like [str] subset str or list of subset. Defaults to ``train``. split : str use (stratified) track splits for validation split (``valid``), defaults to ``train``. seq_duration : float training is performed in chunks of ``seq_duration`` (in seconds, defaults to ``None`` which loads the full audio track samples_per_track : int sets the number of samples, yielded from each track per epoch. Defaults to 64 source_augmentations : list[callables] provide list of augmentation function that take a multi-channel audio file of shape (src, samples) as input and output. Defaults to no-augmentations (input = output) random_track_mix : boolean randomly mixes sources from different tracks to assemble a custom mix. This augmenation is only applied for the train subset. seed : int control randomness of dataset iterations args, kwargs : additional keyword arguments used to add further control for the musdb dataset initialization function. rN)rbrrrrr)musdbrrrrcrrrrrdrDBmusr)r?rrbrrrrrcrrdrrrfkwargsr+s rr@zMUSDBDataset.__init__sl    D (   !2$8! 058       #rcg}d}|jj||jz}|jdkrw|jrot |jjdD]\}}||jkr|}|jrtj |jj}|j|_ tj d|j |jz |_tj|j|jjtj}||}||tj|d}|d} | ||} nt3|jjdd} | || z } ndtj|jjtj} tj|j|jjjtj} | | fS)Nrr$r)dtyper[r)r-rrrrc enumeratesetuprrrrchunk_durationrr chunk_startrH as_tensorr$r7Tfloat32rdrrrrkeysrhtargets) r?rhr target_indtrackkrr7rrrvocinds rrlzMUSDBDataset.__getitem__Bs  )? ?@ : T%6 &tx~i'@AA , , 6T[((!"J(;"M$(/::E(,'8$$*N1entGX6X$Y$Y! f(=(C(EU][[[11%88$$U++++K 1555E ! A%*%dhnY7<<>>??EEhOOf %  U]CCCA dk : @ B%-XXXA!t rcDt|jj|jzSr4)rr-rrros rrpzMUSDBDataset.__len__ts48?##d&<<rYs >>>>>>>>>>>>>>>>  CD6    % 4. @#####     f   EQVQ]    %,    5;#+@f.  #f.+3+=f. <x'9 9:f.f.f.f.RR<R<R<R<R<\R<R<R    >  S"===%%''GD! J !3444)6vt)D)D&M=$ E 9Z9;;<<<  TYuuSS//00 1 1XXQ1171: 0I#II 9 X JOGcc!ff,w6]=V W W W JOGcc!ff,w6]=V W W W E )+BT+IJJJ E %ss='9'9::: E *CC ,>,>???"&MK$//? 0M -((1 agMJr