|h$ddlddlmZddlmZddlmZmZddlm Z ddl m Z ddl ZddlZGdd e jZy) )*)audio)Path)UnionList)nn) perf_counterNceZdZd deeej fdeeefffd ZdejfdZ e de fdZ d dejfd Zd eejfd ZxZS) VoiceEncoderdevice weights_fpathcZt|tjtt t d|_tjt t|_ tj|_ |6tjtjj!rdnd}n%t#|t$rtj|}||_|7t't(j+j,j/d}n t'|}|j1st3d|zt5}tj6|d}|j9|d d |j;||r&t=d |j>t5|z fzyy) a6 If None, defaults to cuda if it is available on your machine, otherwise the model will run on cpu. Outputs are always returned on the cpu, as numpy arrays. :param weights_fpath: path to ".pt" file path. If None, defaults to built-in "pretrained.pt" model T) batch_firstNcudacpuz pretrained.ptz7Couldn't find the voice encoder pretrained model at %s.) map_location model_stateF)strictz5Loaded the voice encoder model on %s in %.2f seconds.) super__init__rLSTMmel_n_channelsmodel_hidden_sizemodel_num_layerslstmLinearmodel_embedding_sizelinearReLUrelutorchr r is_available isinstancestrr__file__resolveparentjoinpathexists Exceptiontimerloadload_state_dicttoprinttype)selfr verboser start checkpoint __class__s X/var/www/html/test/engine/venv/lib/python3.12/site-packages/resemblyzer/voice_encoder.pyrzVoiceEncoder.__init__ sN GGN,=?O]ab ii 13GH GGI  >\\EJJ,C,C,E&5QF  $\\&)F    N224;;DD_UM /M##%U)*+ +ZZ EB  Z 6uE   I;;%01 2 melsc|j|\}\}}|j|j|d}|tj|ddz S)a Computes the embeddings of a batch of utterance spectrograms. :param mels: a batch of mel spectrograms of same duration as a float32 tensor of shape (batch_size, n_frames, n_channels) :return: the embeddings as a float 32 tensor of shape (batch_size, embedding_size). Embeddings are positive and L2-normed, thus they lay in the range [0, 1]. T)dimkeepdim)rr rr!norm)r1r8_hidden embeds_raws r6forwardzVoiceEncoder.forward3sL4;FAYYt{{6":67 EJJzq$GGGr7 n_samplescd|cxkrdksJJtttzdz }ttj|dz|z }ttj t|z |z }d|ksJd|t ksJdt|t zz zgg}}td|t z |zdz}td||D]S} tj| | t zg} | |z} |jt| |jt| U|d} || jz | j| jz z } | |krt|dkDr |dd}|dd}||fS)a" Computes where to split an utterance waveform and its corresponding mel spectrogram to obtain partial utterances of each. Both the waveform and the mel spectrogram slices are returned, so as to make each partial utterance waveform correspond to its spectrogram. The returned ranges may be indexing further than the length of the waveform. It is recommended that you pad the waveform with zeros up to wav_slices[-1].stop. :param n_samples: the number of samples in the waveform :param rate: how many partial utterances should occur per second. Partial utterances must cover the span of the entire utterance, thus the rate should not be lower than the inverse of the duration of a partial utterance. By default, partial utterances are 1.6s long and the minimum rate is thus 0.625. :param min_coverage: when reaching the last partial utterance, it may or may not have enough frames. If at least of are present, then the last partial utterance will be considered by zero-padding the audio. Otherwise, it will be discarded. If there aren't enough frames for one partial utterance, this parameter is ignored so that the function always returns at least one slice. :return: the waveform slices and mel spectrogram slices as lists of array slices. Index respectively the waveform and the mel spectrogram with these slices to obtain the partial utterances. rr;izThe rate is too highz-The rate is too low, it should be %f at leastr:N)int sampling_ratemel_window_stepnpceilroundpartials_n_framesmaxrangearrayappendslicer3stoplen)rCrate min_coveragesamples_per_framen_frames frame_step wav_slices mel_slicesstepsi mel_range wav_rangelast_wav_rangecoverages r6compute_partial_slicesz#VoiceEncoder.compute_partial_slicesBs2<$1$$$$$ !@4!GIrww A 1BBCD=4#7;L"LMN :~555~.. F0_ /2CC D1F F."$RJ Ax"33j@1DEq%, 1A!Q):%:!; of are present, then the last partial utterance will be considered by zero-padding the audio. Otherwise, it will be discarded. If there aren't enough frames for one partial utterance, this parameter is ignored so that the function always returns at least one slice. :return: the embedding as a numpy array of float32 of shape (model_embedding_size,). If is True, the partial utterances as a numpy array of float32 of shape (n_partials, model_embedding_size) and the wav partials as a list of slices will also be returned. r:rconstantNaxis)r`rRrQrHpadrwav_to_mel_spectrogramrNr!no_grad from_numpyr.r rnumpymeanlinalgr>)r1rareturn_partialsrSrTrXrYmax_wave_lengthmelsr8partial_embeds raw_embedembeds r6embed_utterancezVoiceEncoder.embed_utterancews&6"&!)r1rvkwargsrarss r6 embed_speakerzVoiceEncoder.embed_speakersbGG(,.!$2T11#WuWPVW.457 299>>)Q777.sA)NTN)Fg?g?)__name__ __module__ __qualname__rr$r!r rr FloatTensorrB staticmethodrEr`rHndarrayrurry __classcell__)r5s@r6r r s%2uS%,,%67%2[`aegjaj[k%2N HE-- H2&#2&2&h-2::-^ 8$rzz"2 8r7r )resemblyzer.hparams resemblyzerrpathlibrtypingrrr!rtimer r+rkrHModuler r7r6rs/!& f8299f8r7