o if(@stddlmZddlmZmZmZmZmZddlZddl m Z ddl m Z ddl mZddlmZGdd d eZdS) ) OrderedDict)DictListOptionalTupleUnionN) ComplexTensor)DNN_Beamformer)DNN_WPE) AbsSeparatorcGsDeZdZ                        d@dedededededededededededededed ed!ed"ed#ed$ed%ed&ed'ed(ed)ed*ed+ed,ed-ed.ed/ed0ed1ed2ed3ed4efFfd5d6 Z 7dAd8ee j e fd9e j d:e e d;eeee j e fe j effdd?ZZS)BNeuralBeamformermask_mseFblstmp,@Tcrelusigmoid mvdr_soudenHz>ư> input_dimnum_spk loss_typeuse_wpe wnet_typewlayerswunitswprojs wdropout_ratetapsdelayuse_dnn_mask_for_wpe wnonlinearmulti_source_wpewnormalizationuse_beamformer bnet_typeblayersbunitsbprojsbadim ref_channeluse_noise_mask bnonlinearbeamformer_typertf_iterations bdropout_rate shared_powerdiagonal_loading diag_eps_wpe diag_eps_bf mask_flooringflooring_thres_wpeflooring_thres_bfuse_torch_solverc$%sxt||_||_|dvrtd|||_||_|jrf| r#d}$nd}$td$id|d|d|d|d |d | d | d | d | d|rHdn|d| d|$d|d|d|d| d|!d|#|_nd|_||_ |jrt d$id|d|d|d|d|d|d|d|d |d|d|d |d!|d"| d#| d|d|d| d|"d|#|_ nd|_ |o||_ dS)%N)rspectrum spectrum_log magnitudezUnsupported loss type: %sr rwtypewidimr!r"r# dropout_rater%r& use_dnn_masknmask nonlinear iterations normalizationr8diag_epsr;flooring_thresr>bidimbtyper-r.r/rr2r0r1r4r5btapsbdelay) super__init___num_spkr ValueErrorr+rr wper1r beamformerr7)%selfrrrrr r!r"r#r$r%r&r'r(r)r*r+r,r-r.r/r0r1r2r3r4r5r6r7r8r9r:r;r<r=r>rH __class__rP[/home/ubuntu/.local/lib/python3.10/site-packages/espnet2/enh/separator/neural_beamformer.pyrR s )           zNeuralBeamformer.__init__Ninputilens additionalreturnc Csf|dvs J||}t}|jr|jdur|jdr|jrj|dkr:|j|d|\}}| d}n|dkrI|j||\}}|durjt |t rft |j D] }|||d|d<qWn||d <|jr|dkr|j||\}}t |j D] }|||d |d<qt||j kr||j |d <d||fSd} |dkr|jr||d|\}}}} t |t rd d |D}|durt |j D]}d|d} |||| <|| d|d| <qn| d}|dur||d<| d|d <n|jrF|||\}}}} |durFt |t r;t |j D]}d|d} |||| <|||d| <q!n ||d<| d|d <|jr|jjdrk|jjdrk|jrk|jjdkrm|j dkrmd} t |t rwtd|j||| d\}}}t |j D]}|||d |d<qt||j kr||j |d <t |t s|g}|||fS)aForward. Args: input (torch.complex64/ComplexTensor): mixed speech [Batch, Frames, Channel, Freq] ilens (torch.Tensor): input lengths [Batch] additional (Dict or None): other data included in model NOTE: not used in this model Returns: enhanced speech (single-channel): List[torch.complex64/ComplexTensor] output lengths other predcited data: OrderedDict[ 'dereverb1': ComplexTensor(Batch, Frames, Channel, Freq), 'mask_dereverb1': torch.Tensor(Batch, Frames, Channel, Freq), 'mask_noise1': torch.Tensor(Batch, Frames, Channel, Freq), 'mask_spk1': torch.Tensor(Batch, Frames, Channel, Freq), 'mask_spk2': torch.Tensor(Batch, Frames, Channel, Freq), ... 'mask_spkn': torch.Tensor(Batch, Frames, Channel, Freq), ] )rNmaskrr_zmask_dereverb{}r mask_dereverb1z mask_spk{} mask_noise1cSsg|]}|dqS)ra)squeeze).0enhrPrPrZ sz,NeuralBeamformer.forward..z dereverb{}mask_ dereverb1wmpdrwpdzJSingle-source WPE is not supported with beamformer in multi-speaker cases.)powers)dimrtrainingr startswithrrU predict_mask unsqueezerd isinstancelistrangerformatr+rVlenr4r7rFNotImplementedError) rWr[r\r]enhancedothersmask_wspkothers_brlkeyrPrPrZforward}s                     zNeuralBeamformer.forwardcCs|jSN)rS)rWrPrPrZrszNeuralBeamformer.num_spk)"r rFrrrrrrrTrTFTrrrrrrTrrrrTTrrFrrTr)__name__ __module__ __qualname__intstrboolfloatrRrtorchTensorrrrrrrr~propertyr __classcell__rPrPrXrZr s     !"#$%&'t   r ) collectionsrtypingrrrrrrtorch_complex.tensorr!espnet2.enh.layers.dnn_beamformerr espnet2.enh.layers.dnn_wper #espnet2.enh.separator.abs_separatorr r rPrPrPrZs